RU2771955C2 - Device, equipment, method for encoding and decoding video images - Google Patents

Abstract

FIELD: encoding/decoding.

SUBSTANCE: invention relates to means for encoding and decoding video images. The difference of a motion vector of an image affine block is determined. The accuracy of the motion vector of the image affine block is determined. The size of an affine subblock of a motion compensation image in the image affine block is determined based on the difference of the motion vector, the accuracy of the motion vector, the distance between control points in the image affine block and the size of the image affine block. In this case, the size contains a length in the horizontal direction and a length in the vertical direction, wherein the length of the image affine block in the horizontal direction is an integer multiple of the length of the affine subblock of the motion compensation image in the horizontal direction, and the length of the image affine block in the vertical direction is an integer multiple of the length of the affine subblock of the motion compensation image in the vertical direction; and control points, which are pixels used to determine the difference of the motion vector.

EFFECT: increase in the efficiency of encoding video images.

38 cl, 21 dwg, 1 tbl

Description

The technical field to which the invention belongs

The present invention relates to the field of video processing, and more particularly to a video encoding method, a video decoding method, a video encoding equipment, a video decoding equipment, an encoding apparatus, and a decoding apparatus.

State of the art

Due to the rapid development of Internet technology, through which more and more people's material and spiritual needs are met, there is now a growing demand for Internet video applications, especially for high-definition video applications. However, high-definition video has a fairly large amount of data. To transmit high-definition video over the Internet with limited bandwidth, it is necessary, first of all, to solve the technical problem of compressing and encoding high-definition video. Currently, two international organizations are leading the way in improving video coding standards: the Motion Picture Experts Group (MPEG for short) of the International Organization for Standardization (ISO for short)/International Electrotechnical Commission ( International Electrotechnical, “IEC” for short) and the Video Coding Experts Group (“VCEG” for short) of the International Telecommunication Union-Telecommunication Standardization Sector, “ITU-T” for short ). MPEG, founded in 1986, specializes in the development of related standards, which are mainly applied to storage, television broadcast, Internet media streaming or wireless network, and the like in the field of multimedia. ITU-T mainly develops video coding standards for real-time video transmission such as video phones, video conferencing or other applications.

In recent decades, specific video coding standards have been successfully developed internationally for various applications, mainly including: MPEG-1 standard for video compact disc (Video Compact Disc "VCD" for short), MPEG-2 standard for digital video disc (Digital Video Disc, "DVD" for short) and digital television broadcasting (Digital Video Broadcasting, "DVB" for short), H.261 standard, H.263 standard and H.264 videoconferencing standard, MPEG-4 standard, allowing to encode an object in any form, and the latest high efficiency video coding standard (High Efficiency Video Coding, "HEVC" for short).

, где

представляет координаты каждой точки выборки в блоке изображения. Аналогичным образом, изображение может быть также представлено в виде двумерного массива точки выборки и обозначается с помощью способа, аналогичного способу, используемому для блока изображения. Кроме того, точка выборки также может упоминаться как пиксель, и должна быть использована без различия в данном описании настоящего изобретения.According to both the H.264 video coding standard and the latest HEVC video coding standard, which are widely used at present, various types of coding operations such as block prediction, transformation, and entropy coding are performed using an image block as a basic block. The image block is a two-dimensional sample point array, which can be a square array or a rectangular array. For example, a 4x4 image block can be considered as a square sample point array including a total of 4×4, ie 16 sample points. As shown in Fig. 1, the image block may be denoted as B. The image block signal is the sample value of the sample point in the image block, and may be represented as

, where

represents the coordinates of each sample point in the image block. Similarly, an image may also be represented as a two-dimensional sample point array and denoted in a manner similar to that used for an image block. In addition, a sampling point may also be referred to as a pixel, and should be used without distinction in this description of the present invention.

The video sequence includes images arranged frame by frame. As a rule, successive images are generated in a fairly short time interval, and the images do not change much. Therefore, there is little difference between successive images. Thus, only the difference between the current picture and the reference picture can be encoded based on the temporal correlation between successive pictures. The reference picture is typically a reconstructed encoded picture whose generation time is close to the current picture. Thus, the amount of information to be encoded is reduced and a signal compression effect is achieved. In video coding technology, this type of method is referred to as inter-picture prediction coding technology, and the technology is widely used in modern video coding standards. Fig. 2 shows an inter prediction coding architecture. The image to be encoded, which is usually a frame-by-frame video image, is input and divided into image blocks that do not overlap with each other. The encoder obtains a motion vector of the image block based on the motion estimate, and the encoder obtains, by motion compensation, a motion compensation prediction signal of the image block using the motion vector of the image block. Then, the motion-compensated prediction signal is subtracted from the original image block signal to obtain a prediction residual signal. After undergoing transformation, quantization, and entropy encoding, the prediction residual signal is written to the bit stream. In the encoding process, if the received picture block motion compensation prediction signal is more accurate, the number of obtained prediction residual signals is smaller, the number of bits to be encoded is smaller, and the compression ratio of the picture block is higher.

изображения, блока

изображения, который имеет такой же размер, и который наилучшим образом соответствует текущему блоку изображения. Смещение местоположения блока

изображения относительно

в пространстве, называют вектором движения блока

изображения. Компенсация движения представляет собой процесс получения сигнала предсказания из опорного изображения на основании вектора движения блока

изображения. Полученный сигнал упоминается как сигнал предсказания с компенсацией движения.To obtain a motion estimate from a reference image using the current block

images, block

an image that is the same size and that best fits the current image block. Block location offset

images regarding

in space is called the motion vector of the block

Images. Motion compensation is the process of deriving a prediction signal from a reference image based on the motion vector of the block

Images. The received signal is referred to as a motion-compensated prediction signal.

The decoder side performs de-quantization and inverse transformation on the prediction residual signal of the picture block to obtain the reconstructed residual signal, and then obtains the motion compensation prediction signal from the reference picture by motion compensation prediction based on the motion vector of the picture block, and can obtain the recovered signal. of the image block by summing the motion compensation prediction signal and the reconstructed image block residual signal.

в момент

делят на блоки изображения, которые имеют различные размеры и которые не перекрываются друг с другом. Стрелка в каждом блоке изображения схематически представляет вектор движения, определяемый стороной кодера, и длина, и направление, соответственно, представляют интенсивность движения и направление движения текущего блока изображения. Вектор движения каждого блока изображения применяется ко всем точкам выборки в блоке изображения. Более конкретно, сигнал предсказания каждого блока изображения получают из указанного опорного изображения для всех точек выборки в блоке изображения на основе смещения местоположения, представленного вектором движения.3 shows the division of the image block and the relationship between the image block and the motion vector. Image

in the moment

divide into image blocks that have different sizes and that do not overlap with each other. An arrow in each image block schematically represents a motion vector defined by the encoder side, and the length and direction respectively represent the motion intensity and motion direction of the current image block. The motion vector of each image block is applied to all sample points in the image block. More specifically, the prediction signal of each image block is obtained from the specified reference image for all sample points in the image block based on the location offset represented by the motion vector.

Motion-compensated prediction based on the block structure is more suitable for implementation in software and hardware, and has low complexity, and can achieve relatively high compression efficiency while limiting computing resources. However, there are also some disadvantages. So, for example, in the method, one motion vector is used to indicate the motion of all pixels in an image block. When all pixels have the same direction of motion and the same size, such as in translational motion, the method can be used to accurately indicate the motion of all pixels. When pixels in an image block have different directions of movement or sizes, such as in movements such as zoom and rotate, the method cannot be used to accurately indicate the movement of all pixels. Therefore, an accurate prediction signal cannot be obtained, and compression performance is degraded.

To eliminate the shortcomings of the existing motion-compensated prediction technology, a motion-compensated prediction technology based on a linearly varying motion vector field is proposed. The motion vector field is the motion vector field of the image block. The image block is divided into smaller sub-blocks. A subblock can be as small as a pixel and as large as the original block. Each subblock has a motion vector. The motion vectors of all sub-blocks in the image block form the motion vector field of the image block. The motion vector field of the image block may vary linearly or non-linearly. The fact that the motion vector field of the image block changes linearly means that the motion vectors of neighboring sub-blocks in the image block change uniformly. If the motion vectors of neighboring blocks in the image change block change unevenly, then the motion vector field of the image block changes non-linearly. This specification is mainly focused on motion compensation prediction technology based on a motion vector field that varies linearly. In a paper titled "Obtaining the motion vector of a deformable block" (Image Processing (ICIP), 2012 19 ^th IEEE International Conference, pp.1549-1552, September 30 2012 to October 3 2012) proposed by Na Zhang, Xiaopeng Fan, Debin Zhao and Wen Gao, which uses four vertices in the image block as control points, and performs a bilinear interpolation method on the motion vectors of the four control points to obtain the motion vector of each 4x4 block in the image block. The motion vectors of the four vertices in the image block are derived from the motion vectors of the surrounding image blocks, and the motion vectors of all 4x4 blocks form the motion vector field of the image block. Then, a motion-compensated prediction signal of the image block is obtained from the reference image based on the motion vector field. In the motion compensation prediction technology based on the motion vector field, the size of a block having an independent motion vector is reduced, non-translational motions such as rotation and zoom can be specified more accurately, and a more accurate motion compensation prediction signal can be obtained. The above motion vector field is obtained by performing bilinear interpolation of motion vectors of control points, and the motion vector field satisfies the bilinear model. Bilinear interpolation is a mathematical way of interpolation. With reference to Fig. 4, the following is a description of a specific interpolation method:

в

и значения функции

в четырех точках

,

,

и

уже известны. Во-первых, выполняют линейную интерполяцию в направлении

, чтобы получить:It is assumed that the value of the function should be calculated

in

and function values

at four points

,

,

and

are already known. First, linear interpolation is performed in the direction

, To obtain:

(1)

(one)

, чтобы получить:Then, linear interpolation is performed in the direction

, To obtain:

(2)

(2)

может быть получено путем подстановки результата, полученного из формулы (1) в формулу (2). Векторы движения всех точек в блоке изображения могут быть получены на основании векторов движения контрольных точек, используя способ интерполяции.Point value

can be obtained by substituting the result obtained from formula (1) into formula (2). The motion vectors of all points in an image block can be derived from the motion vectors of control points using an interpolation method.

In this method, the motion vector of the reference point is selected based on the motion vector of the surrounding image block, and the selected motion vector may not be accurate. If the cue motion vector is obtained by motion estimation and searching, the cue motion vector must be written to the bit stream. In this case, a more efficient method is required, such as using an affine transformation model to specify the motion vector field of the image block. Thus, a motion-compensated prediction technology based on affine transformation has been proposed.

являются координатами пикселя

аффинного блока изображения в текущем изображении,

являются координатами пикселя

, который находится в опорном изображении, и соответствует пикселю

, и

являются параметрами аффинного преобразования. Если параметры аффинного преобразования известны, то может быть вычислена позиция

пикселя

в опорном изображении и, следовательно, может быть получен сигнал предсказания пикселя из опорного изображения.An affine transformation is a mapping relationship between two spaces, and a general affine transformation includes rotation, scaling, and translation. The motion vector field, which represents the image block, and which is obtained using the affine transformation model, changes linearly, and the motion vector field satisfies the affine transformation. The general 6-parameter affine transformation model can represent the above three movements. In particular, as shown in formula (3),

are pixel coordinates

affine image block in the current image,

are pixel coordinates

, which is in the reference image, and corresponds to a pixel

, and

are parameters of the affine transformation. If the parameters of the affine transformation are known, then the position can be calculated

pixel

in the reference image, and therefore a pixel prediction signal can be obtained from the reference image.

(3)

(3)

The affine parameter is usually calculated based on the motion vectors of some pixels in the image block. The motion vector can be expressed by formula (4):

(4)

(4)

Formula (3) is substituted into formula (4) to get:

(5)

(5)

, и вектор движения является

. Координаты верхней правой угловой точки являются

, и вектор движения является

. Координаты левой нижней угловой точки составляют

, и вектор движения равен

. Может быть получена следующая модель аффинного преобразования 6-параметров путем раздельных подстановок трех векторов

,

и

движения в формулу 5:Specific values of the affine transformation parameters can be obtained by substituting a plurality of groups of known motion vectors, for example, three groups of known motion vectors, into formula (5). Thus, the motion vectors of all points in the image block can be calculated using formula (3) and formula (4). In practice, generally, the motion vector of a corner point (reference point) in an image block is first obtained by motion estimation, then an affine transform parameter is calculated based on the corner point motion vector, and the motion vectors of all pixels in the image block are calculated. In FIG. 5(a) and FIG. 5(b) shows the motion vector of the corner point. Fig. 5(a) illustrates an image block X. The width of the image block is w height h. The coordinates of the top left corner point are

, and the motion vector is

. The coordinates of the top right corner point are

, and the motion vector is

. The coordinates of the lower left corner point are

, and the motion vector is

. The following model of affine transformation of 6-parameters can be obtained by separate substitutions of three vectors

,

and

movements in formula 5:

(6)

(6)

The motion vectors of the three vertices in the image block may differ. As shown in Fig. 5(b), the image block ratio display in the reference image is indicated by dotted lines. After calculating the motion vectors of all points in the image block, motion-compensated prediction signals of all points in the image block are obtained by motion-compensated prediction.

The content above uses the 6-parameter affine transformation model as an example. However, the affine transformation model is not limited to the 6-parameter model, and may be the 4-parameter affine transformation model shown in formulas (7) and (8), the 2-parameter scaling model shown in formula (9), or the like. P. If a 4-parameter affine transform model is selected for an image block for motion-compensated prediction, the affine model has four unknown parameters, and the model parameters can be obtained using two groups of motion vectors. The two groups of motion vectors may be the motion vectors of two corner points in the image block, or may be the motion vectors of two points in the image block.

(7)

(7)

(8)

(eight)

(9)

(nine)

Formula (3) is substituted into formula (7) to get:

(10)

(ten)

, и вектор движения представляет собой

. Координаты верхнего правого угла являются точки

, и вектор движения представляет собой

. Координаты и векторы движения обоих угловых точек подставляют в формулу (10), и может быть получена формула модели движения, для которых верхняя левая угловая точка и верхняя правая угловая точка используют как контрольные точки:The coordinates of the upper left corner are points

, and the motion vector is

. The coordinates of the top right corner are points

, and the motion vector is

. The coordinates and motion vectors of both corner points are substituted into formula (10), and a motion model formula can be obtained for which the upper left corner point and the upper right corner point are used as control points:

(11)

(eleven)

, и вектор движения представляет собой

. Координаты левой нижней угловой точки являются

, и вектор движения представляет собой

. Координаты и векторы движения обоих угловых точек подставляют в формулу (10), и получают формулу модели движения, для которой верхняя левая угловая точка и нижняя левая угловая точка используются как контрольные точки:The coordinates of the top left corner point are

, and the motion vector is

. The coordinates of the lower left corner point are

, and the motion vector is

. The coordinates and motion vectors of both corner points are substituted into formula (10), and a motion model formula is obtained, for which the top left corner point and bottom left corner point are used as control points:

(12)

(12)

Compared with the latest HEVC video coding standard, for a sequence of images/video frames including motion rotation and scaling, motion compensation prediction technology based on a motion vector field that varies linearly can greatly improve coding performance. A motion vector field that varies linearly includes a motion vector field generated using an affine transformation model and a motion vector field generated using a bilinear model. Motion compensation prediction technology based on affine transform is used as an example. After obtaining the affine transform parameter of the image block, the motion vector of each dot in the image block must be calculated, and motion-compensated prediction is performed based on the motion vector of each dot to obtain a motion-compensated prediction signal of each dot. Since the motion vectors of all pixels in an image block may differ, various operations must be performed on the pixels based on the motion vectors. Pixel-based motion compensation prediction has a high complexity. To reduce the complexity of encoding/decoding in the related technology, an image block must be further divided into sub-image blocks, a motion vector of each image sub-block is calculated, and then a motion-compensated prediction signal of each image sub-block is obtained.

However, in the related technology, the sub-block sizes of an image are fixed. An excessively small size image sub-block results in a relatively high level of encoding/decoding complexity, and an excessively large image sub-block reduces encoding/decoding performance. Therefore, in practice, the application of such an efficient coding method is largely limited.

Disclosure of the essence of the invention

The present invention relates to a video decoding method, a video encoding method, a decoding apparatus, an encoding apparatus, a decoding equipment, and an encoding equipment. An image sub-block with an appropriate size can be selected so that the entire image block is divided into a plurality of image sub-blocks with the same size for processing in order to reduce encoding/decoding complexity and also improve encoding/decoding performance.

According to a first aspect, a video image decoding method is provided, including: determining a motion vector difference of an affine image block; determining the accuracy of the motion vector of the affine image block; determining the size of an affine image sub-block of a motion compensation image in an affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, where the size includes a length in the horizontal direction and a length in the vertical direction, so that the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image block of the horizontal direction motion compensation image subblock, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation image in the vertical direction; and control points are pixels used to determine the motion vector difference; and performing decoding processing on the affine image block based on the size of the affine subblock of the motion compensation image.

In the video image decoding method presented in accordance with the first aspect of the present invention, the size of an affine sub-block of a motion compensation image is adaptively determined based on the determined two parameters, i.e., the motion vector difference and the motion vector accuracy, and in combination with the distance between control points in the affine block image and the size of the affine image block, so that the affine image block can be divided into a plurality of affine motion compensation image sub-blocks with the same size. This can greatly reduce the complexity on the decoder side and improve decoding efficiency.

Based on the first aspect of the present invention, there is provided a method for determining and adjusting the size of an affine sub-block of a motion compensation image, which may be specifically: proportionally adjusting the first horizontal/vertical distance based on the proportion of the motion vector accuracy in the first component/second component of the motion vector difference to obtain the length of the affine a horizontal/vertical motion compensation image sub-block; and determining whether the length of the affine image block in the horizontal/vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation in the horizontal/vertical direction and, if the length of the affine image block in the horizontal/vertical direction is not an integer multiple of the length of the affine subblock of the horizontal/vertical motion compensation image by adjusting the length of the horizontal/vertical motion compensation image affine sub-frame so that the horizontal/vertical direction affine image block length is an integer multiple of the adjusted length of the horizontal/vertical motion compensation affine sub-image direction. In this particular implementation, the length of the horizontal/vertical motion compensation affine sub-image subblock is obtained by adjusting the first horizontal/vertical distance between control points based on the proportion of the relationship between the motion vector accuracy and the first component/second component of the motion vector difference, so that the size of the affine sub-image block, the motion compensation may match the motion vector accuracy of the affine image block. This ensures the quality of the decoding and reduces the complexity of the decoding.

Based on the first aspect of the present invention, adjusting the length of the horizontal/vertical direction motion compensation affine image subblock such that the length of the horizontal/vertical direction affine image block is an integer multiple of the corrected length of the horizontal/vertical motion compensation affine image subblock can be specifically: adjusting, i.e., increasing/decreasing, the length of the motion compensation affine sub-image block in the horizontal/vertical direction by one or more units of length, and then cycling the above decision until the length of the affine image block in the horizontal/vertical direction will not be an integer multiple of the adjusted length of the horizontal/vertical motion compensation affine sub-image subblock. The length unit is typically one block of pixels. More specifically, the length of the horizontal/vertical motion compensation image affine sub-frame is corrected, that is, increased/decreased by 1. In addition, the above correction may be a correction, that is, the length of the horizontal/vertical motion compensation image affine sub-frame is increased/decreased by a fixed step size, and the step size is a correction value, that is, a basic unit used to correct the length of the affine motion compensation image subblock in the horizontal/vertical direction in each cycle. The step size can be any value and is typically one pixel block, or it can be another value such as 2.5 pixel blocks. It is possible that the value of the step size/correction used in the above correction may change depending on the ratio of the original length (ie, uncorrected length) of the horizontal/vertical direction motion compensation affine image subblock. For example, when the initial length of the affine subblock of the horizontal direction motion compensation image is an odd number, an odd step size is selected. For example, an odd number of length units, such as one length unit, is used as the step size for correction. When the initial length of the affine subblock of the horizontal motion compensation image is an even number, an even step size is selected. For example, an even number of length units is used as the step size. In this embodiment, it can be ensured that the length of the horizontal/vertical motion compensation affine sub-image block can be rapidly approached to a value by which the length of the horizontal/vertical direction affine image block is divided and the complexity is further reduced.

In order to improve efficiency and reduce running time, in the decoding method presented in accordance with the first aspect of the present invention, the length of the affine sub-image block of the motion compensation in the horizontal/vertical direction can be corrected to the nearest value by which the length of the affine image block in the horizontal/vertical direction direction is divided. If the difference between the length of the horizontal/vertical direction motion compensation affine sub-image block and two adjacent values by which the horizontal/vertical direction affine image block length is divided are equal, then the correction can be selected according to the requirements of the encoder side and the decoder side. For example, if a low latency is required, then an adjustment, i.e. an increase, may be chosen. If high quality is required, an adjustment, i.e. reduction, can be chosen.

, W_max - это возможно максимальная длина аффинного блока изображения в горизонтальном/вертикальном направлении. Определяют интервал, в течение которого длина аффинного субблока изображения компенсации движения в горизонтальном/вертикальном направлении снижается, и значение нижнего предела или значение верхнего предела интервала, в течение которого длина аффинного субблока изображения компенсации движения в горизонтальном/вертикальном направлении снижается, используются как окончательно определенная длина аффинного субблока изображения компенсации движения в горизонтальном/вертикальном направлении. Для правила получения значения нижнего предела или значения верхнего предела, сделана ссылка к соглашению между стороной кодера и стороной декодера. Верхнее предельное значение или нижнее предельное значение приближенного интервала может быть явно указано, используя битовый поток или может быть выбрано при округлении. С помощью указанной выше таблицы, конечная длина аффинного субблока изображения компенсации движения в горизонтальном/вертикальном направлении может быть определена с помощью таблицы поиска на основе исходной длины аффинного субблока изображения компенсации движения в горизонтальном/вертикальном направлении. Таким образом, сложность системы снижается, и время вычисления уменьшается.When it is determined that the length of the affine image block in the horizontal/vertical direction is not an integer multiple of the length of the affine sub-image block of the horizontal/vertical motion compensation, in order to further simplify the algorithm, the same table on the encoder side can be separately set and side of the decoder, and the possible maximum length of the affine image block in the horizontal/vertical direction is divided into a plurality of intervals, using the allowable minimum divisor as the initial value, for example, 2 ^N . The division interval is 2 ^N+i , where

, W _max is the possible maximum length of an affine image block in the horizontal/vertical direction. The interval during which the length of the horizontal/vertical motion compensation image affine sub-frame decreases, and the lower limit value or the upper limit value of the interval during which the length of the horizontal/vertical motion compensation image affine sub-frame decreases are used as the finally determined length. affine sub-image compensation motion in the horizontal/vertical direction. For the rule for obtaining a lower limit value or an upper limit value, reference is made to an agreement between the encoder side and the decoder side. The upper limit value or the lower limit value of the approximate interval may be explicitly specified using the bitstream or may be chosen by rounding. With the above table, the final length of the horizontal/vertical motion compensation image affine sub-frame can be determined with a lookup table based on the initial length of the horizontal/vertical motion compensation image affine sub-frame. Thus, the complexity of the system is reduced and the computation time is reduced.

Alternatively, to further reduce the computational complexity in the solution, in the decoding method presented in accordance with the first aspect of the present invention, not only if the length of the affine image block in the horizontal/vertical direction is an integer multiple of the length of the affine sub-image block of the motion compensation image in horizontal/vertical direction, but also determines whether the length of the affine subblock of the horizontal/vertical motion compensation image is an integer multiple of the predetermined length S, which facilitates the calculation. More specifically, it should be determined whether the length of the affine subblock of the horizontal/vertical motion compensation image satisfies the following two conditions:

(1) the length of the affine image block in the horizontal/vertical direction is an integer multiple of the length of the affine sub-image block of the horizontal/vertical motion compensation; and

(2) the length of the affine subblock of the horizontal/vertical motion compensation image is an integer multiple of the predetermined length S.

If the length of the horizontal/vertical direction motion compensation image affine subblock does not satisfy the two conditions, then the length of the horizontal/vertical direction motion compensation image affine subblock is adjusted so that the length of the horizontal/vertical direction motion compensation image affine subblock satisfies condition (1), and condition (2). More specifically, the horizontal/vertical direction affine image block length is an integer multiple of the corrected length of the horizontal/vertical motion compensation affine image subblock, and the adjusted length of the horizontal/vertical motion compensation affine image subblock is an integer multiple of given length. The predetermined length S is typically 4, or may be a value such as 8, 16, or 32, or S is 2 ⁿ , where n is zero or a positive integer. It should be noted that the value of S should not be larger than the size of the affine image block. This solution is applicable to an affine image block in any form. When the affine image block is divided according to the Quad Tree rule, or the size of the affine image block is 2Nx2N, namely in the form of a square, provided that any of the above conditions (1) and (2) is satisfied, the other condition can be met. Thus, only one of the defining conditions must be activated/applied. The sequence for determining whether the conditions (1) and (2) are met and the sequence for adjusting the length of the horizontal/vertical motion compensation affine subblock image based on the determination result may be randomly interchanged.

In the decoding method presented in accordance with the first aspect of the present invention, in order to avoid the case in which the size of the affine sub-image sub-block of the motion compensation image is invalid, that is, less than the allowable minimum size but larger than the size of the affine image block, can be selected and the operation of limiting the amplitude of the length, determined in the above way, of the affine sub-block of the image, motion compensation in the horizontal/vertical direction, is performed. More specifically, a comparison is used to ensure the length of the horizontal/vertical motion compensation image affine subblock to be in the range from the lower limit Tm/Tn to the upper limit W/H. Tm/Tn is a predetermined value and is typically 4, and W/H is the length of the affine image block in the horizontal/vertical direction.

In addition, the decoding method provided in the first aspect of the present invention is applicable to an affine transformation model based on three control points and an affine transformation model based on two control points. In an affine transformation model based on three control points, the lengths of the affine sub-image compensation image in the vertical direction and in the horizontal direction must be independently obtained and independently adjusted. Accordingly, in the affine transformation model based on two control points, only a length that is the length of the affine sub-image subblock of the motion compensation image in one of the horizontal direction and the vertical direction, and which is to be obtained independently from each other, is obtained, and the length in the other direction is additionally receive; and lengths in both directions must be adjusted independently. When the size of the affine image block is 2Nx2N right square, since the lengths of the affine image block in the horizontal direction and the vertical direction are the same, only one of the lengths of the affine image block in the horizontal direction and the vertical direction needs to be determined and adjusted.

According to a second aspect, there is provided a method for encoding a video image, which mainly comprises: determining a motion vector difference of an affine image block; determining the accuracy of the motion vector of the affine image block; determining the size of an affine subblock of a motion compensation image in an affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, the size of the affine image block, where the size includes the length in the horizontal direction and the length in the vertical direction, so that the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image subblock of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation in the vertical direction; and control points are pixels used to determine the motion vector difference; and performing encoding processing on the affine image block based on the size of the affine subblock of the motion compensation image.

In the video encoding method presented in the second aspect of the present invention, the size of an affine sub-block of a motion compensation image is adaptively determined based on two parameters determined, i.e., the motion vector difference and the motion vector accuracy, and in combination with the distance between breakpoints in the affine image block and the size of the affine image, so that the affine image block can be divided into a plurality of affine motion compensation image sub-blocks of the same size. This can greatly reduce the complexity on the encoder side and improve the coding efficiency. The video encoding method according to the second aspect of the present invention includes the same main steps, that is, determining the size of the motion compensation image affine subblock and adjusting the size of the motion compensation image affine subblock, as the video image decoding method according to the first aspect of the present invention, but includes a different last step. Therefore, the solutions in the video decoding method according to the first aspect of the present invention are also applicable to the video encoding method according to the second aspect of the present invention. For more information, see the previous description of the video decoding method in the first aspect.

According to a third aspect, there is provided equipment for decoding a video image, mainly including: a motion vector difference determination unit, configured to determine a motion vector difference of an affine image block; a motion vector accuracy determination module, configured to determine the motion vector accuracy of the affine image block; an image sub-block size determination module configured to determine the size of an affine image sub-block of a motion compensation image in an affine image block based on a motion vector difference, motion vector accuracy, a distance between control points in the affine image block, and a size of the affine image block, where the size includes a length in the horizontal direction and a length in the vertical direction, such that the length of the affine image block in the horizontal direction is an integer multiple of the length of the motion compensation affine image sub-image in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine a vertical motion compensation image sub-block; and control points are pixels used to determine the motion vector difference; and a decoding unit configured to perform decoding processing on the affine image block based on the size of the affine sub-block of the motion compensation image.

The video decoding equipment provided in the third aspect of the present invention corresponds to the video decoding method presented in the first aspect of the present invention, has the function of performing all the steps in the video decoding method provided in the first aspect of the present invention, and the same technical effects, that is, greatly reduce the complexity on the decoder side and improve decoding efficiency.

According to a fourth aspect, there is provided equipment for encoding a video image, mainly including: a motion vector difference determination unit, configured to determine a motion vector difference of an affine image block; a motion vector accuracy determination module, configured to determine the motion vector accuracy of the affine image block; an image sub-block size determination module configured to determine the size of an affine image sub-block of a motion compensation image in an affine image block based on a motion vector difference, motion vector accuracy, a distance between control points in the affine image block, and a size of the affine image block, where the size includes a length in the horizontal direction and a length in the vertical direction, such that the length of the affine image block in the horizontal direction is an integer multiple of the length of the motion compensation affine image sub-image in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine a vertical motion compensation image sub-block; and control points are pixels used to determine the motion vector difference; and an encoding unit configured to perform encoding processing on the affine image block based on the size of the affine sub-block of the motion compensation image.

The video encoding equipment provided in the fourth aspect of the present invention corresponds to the video encoding method presented in the second aspect of the present invention, and has the function of realizing all the steps of the video encoding method presented in the second aspect of the present invention, and the same technical effects can be obtained that is, greatly reduce the complexity on the encoder side and improve the coding efficiency.

According to a fifth aspect, there is provided a video decoding apparatus including a processor and a read-only memory, where the processor calls a program code stored in the memory to implement the decoding method presented in accordance with the first aspect of the present invention.

According to a sixth aspect, there is provided a video encoding apparatus including a processor and a read-only memory, where the processor calls a program code stored in the memory to implement the encoding method presented in accordance with the second aspect.

Based on the above technical solutions, in accordance with the video encoding method, the video decoding method, the encoding equipment, the decoding equipment, the encoding apparatus and the decoding apparatus in the embodiments of the present invention, an affine sub-block size of a motion compensation image is adaptively determined based on determined two parameters, that is, the difference of the motion vector and the motion vector accuracy, and in combination with the distance between the control points in the affine image block and the size of the affine image, so that the affine image block can be divided into a plurality of affine motion compensation image sub-blocks with the same size. This reduces the complexity of encoding/decoding and improves the efficiency of encoding/decoding.

Brief description of the drawings

In order to more clearly describe the technical solutions in the embodiments of the present invention, the following brief description of the accompanying drawings is given, necessary for describing the embodiments. Obviously, the accompanying drawings in the following descriptions show only some embodiments of the present invention, and a person skilled in the art can still obtain other drawings from these attached drawings without creative effort.

1 is a block diagram of an image in a video encoding process;

Fig. 2 is a diagram of a basic video encoding procedure;

Fig. 3 is a picture block division scheme and a relationship between an image block and a motion vector in a video encoding process;

Fig. 4 is a diagram of obtaining a motion vector field of an image block using bilinear interpolation;

Fig. 5(a) and Fig. 5(b) are diagrams for obtaining a motion vector field of an image block using an affine transformation model;

Fig. 6 is a structure diagram of a video decoder used for a decoding method according to an embodiment of the present invention;

7 is a flowchart of a video decoding method according to an embodiment of the present invention;

Fig. 8 is a diagram of an affine image block and control points according to an embodiment of the present invention;

Fig. 9 is another diagram of an affine image block and control points according to an embodiment of the present invention;

Fig. 10 is a schematic representation of three non-overlapping pixels in an affine image block according to an embodiment of the present invention;

11 is another diagram of an affine image block according to an embodiment of the present invention;

12 is another flowchart of a video decoding method according to an embodiment of the present invention;

13 is another flowchart of a video decoding method according to an embodiment of the present invention;

14 is a flowchart of a video encoding method according to an embodiment of the present invention;

15 is another flowchart of a video encoding method according to an embodiment of the present invention;

16 is another flowchart of a video encoding method according to an embodiment of the present invention;

Fig. 17 is a block diagram of a decoding apparatus according to an embodiment of the present invention;

Fig. 18 is a block diagram of an encoding apparatus according to an embodiment of the present invention;

Fig. 19 is a block diagram of a decoding apparatus according to an embodiment of the present invention; and

20 is a block diagram of an encoder according to an embodiment of the present invention.

Implementation of the invention

It is an object of the present invention to reduce the complexity of encoding/decoding while maintaining the performance of motion-compensated prediction based on affine transform, so that the method can be efficiently applied in practice. In this description, an image block that independently has an affine transformation parameter is called an affine image block, that is, an affine block. A sub-image that is obtained by dividing an affine image block and used for motion compensation prediction is referred to as an affine sub-image of motion compensation (Affine Motion Compensation, "Affine-MC" for short) and is referred to as an affine MS block. The technical solution presented in the present invention is based on the adaptive division of an affine block into a plurality of affine MC blocks with the same size, in accordance with the encoding/decoding requirement. The specific method is to adaptively divide, based on the accuracy of the motion vector of the affine block, the motion vector of the affine block, and the distance between control points in the affine block, the affine block into affine MC blocks that satisfy the requirements.

Next, the decoder side is used as an example to describe the solutions of the present invention with specific embodiments.

On the decoder side, a decoder is generally used to implement a decoding method, and decodes and reconstructs the encoded video bitstream. The basic logical structure of the decoder is shown in FIG. 6. The video decoder 60 includes an entropy decoding block 650, a prediction block 652, a dequantization block 654, an inverse transform block 656, a reconstruction block 658, a filter block 659, and a decoded image buffer 660. The prediction block 652 includes a motion compensation block 662 and an intra prediction block 664 . The entropy decoding block 650 includes a normal CABAC encoding/decoding block 666 and a bypass encoding/decoding block 668. In another embodiment, video decoder 60 may include more or less or different functional components.

The basic workflow of video decoder 60 is as follows: upon receiving an encoded video bitstream (referred to as a bitstream for short), entropy decoder 650 parses the bitstream to extract a syntax element from the bitstream. As part of bitstream parsing, entropy decoder 650 may parse a syntax element that is in the bitstream and that has been entropy encoded. The prediction block 652, the dequantization block 654, the inverse transform block 656, the reconstruction block 658, and the filter block 659 can decode video data based on the syntax element extracted from the bitstream, that is, generate decoded video data. The syntax element may include a conventional CABAC encoded/decoded binary codeword and a bypass encoded/decoded binary codeword. Entropy decoder 650 may use conventional CABAC encoder/decoder 666 to decode the conventional CABAC encoded/decoded binary codeword, and may use bypass encoder/decoder 668 to decode the bypass encoded/decoded binary codeword.

The prediction block 652 includes a motion compensation block 662 and an intra prediction block 664 . If inter-picture prediction encoding is used for the image block to be decoded, then the motion compensation unit 662 may determine one or more reference blocks based on motion information of the image block to be decoded, and generate a predicted pixel block of the image block to be decoded based on the reference block. . When intra prediction encoding is used for an image block to be decoded, intra prediction block 664 performs intra prediction to generate a predicted pixel image block. An intra prediction block 664 may generate a predicted pixel block of an image block to be decoded based on a pixel of a spatial neighboring image block using the intra prediction mode.

The recovery section 658 recovers the predicted pixel block obtained by the prediction section 652 and the residual data of the image block to be decoded, which were processed by the dequantization section 654 and the inverse transform section 656, to obtain the reconstructed image block of the image block to be decoded. An image frame that includes a plurality of image blocks is filtered by a filter block 659 and then stored in a decoded image buffer 660 . Thus, the decoding of the encoded video bitstream is completed. The decoded image stored in the buffer 660 of the decoded image 660 may be directly output to a display device for display, or may be used as a reference image in a subsequent bitstream decoding process.

The technical solution of the present invention is applied to a process in which the motion compensation block 662 performs motion compensation predictive decoding. An example of the decoding process is shown in Fig.7. It should be noted that the decoding method and the encoding method in the present invention are essentially the same, and the decoding method is the reverse process of the encoding method. Although the decoding method is used as an example in the following, unless explicitly stated otherwise, the method is basically the same as the encoding method. Thus, in the decoding process described in the following, the explanation and description associated with encoding does not represent any conflict or inconsistency with the solution.

7 is a schematic flowchart of a video decoding method according to an embodiment of the present invention. Specifically, as shown in FIG. 7, method 700 includes the following steps:

S710. Determine the difference of the motion vector of an affine image block.

S720. Determine the accuracy of the motion vector of an affine image block.

S730. Determine the size of an affine image sub-block of a motion compensation image in an affine image block based on the motion vector difference, the motion vector precision, the distance between control points in the affine image block, and the size of the affine image block, where the size includes the length in the horizontal direction and the length in the vertical direction, so that the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image subblock of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation in the vertical direction; and control points are pixels used to determine the motion vector difference.

S740. Perform decoding processing on the affine image block based on the size of the affine subblock of the motion compensation image.

Specifically, the decoding/encoding apparatus determines the motion vector difference of the affine image block based on the determined reference points, determines the motion vector accuracy of the affine image block, and determines the size of the affine sub-image sub-block of the motion compensation in the affine image block based on the determined motion vector difference and the motion vector accuracy. , a distance between control points, and an affine image block size, where the size of the motion compensation affine image subblock is the size used to divide the affine image block, and the affine image block is divided into a plurality of affine motion compensation image subblocks with the same size based on the division size. Then, the decoder/encoder decodes the affine image block based on the affine sub-blocks of the motion compensation image. It should be noted that the affine image block, as a rule, has the shape of a regular rectangle or square. Accordingly, the motion compensation image affine subblock also has the shape of a regular rectangle or square. Thus, after determining the length of the motion compensation affine sub-image in the horizontal direction and the vertical direction, the total size of the motion compensation affine sub-image can be determined. The horizontal direction and the vertical direction are a coordinate system that is commonly used in the prior art and is used to find the position of an image block/affine image block in a video image. The horizontal direction is generally the long side direction of the conventional rectangular video image, the vertical direction is generally the short side direction of the conventional rectangular video image, and the horizontal direction and the vertical direction are perpendicular to each other. Without loss of generality, the horizontal direction may be replaced with a first direction, and the vertical direction may be replaced with a second direction perpendicular to the first direction. However, for ease of description, the horizontal direction and the vertical direction are used for description in this specification. In addition, "affine image block decoding" described in the decoding method herein is a process of parsing encoded bitstream data and reconstructing an encoded image using an affine image block as a base block. Although "affine image block decoding" is used, the affine image block cannot be considered to have been recovered.

When executing the video image decoding method in this embodiment of the present invention, the motion compensation block 662 determines the size of the motion compensation image affine sub-block based on the determined difference of the motion vector and the motion vector precision of the affine image block, the distance between control points, and the size of the affine image block, divides the affine block image into a plurality of affine motion compensation sub-images of the same size based on the size of the affine sub-image of the motion compensation image, and then performs decoding processing. Thus, in the process of decoding an affine sub-block of a motion compensation image with an appropriate size can be adaptively selected based on the change of the above parameters, the decoding complexity is reduced and the performance is improved.

It should be understood that in this embodiment of the present invention, an affine image block (Affine) is an image block whose motion vectors of all blocks correspond to the same affine models, for example, the same affine model and can be represented using the same parameter or the same parameter groups. A block is a group of pixels, for example, may be a pixel, or may be a 1×3 or 2×4 block structure or the like. This is not limited in the present invention. In addition, in this embodiment of the present invention, any non-overlapping pixels can be selected as reference points for determining the motion vector difference of the affine block. This is not limited in the present invention.

Preferably, S710 includes:

determining a first motion vector difference component based on a difference between the motion vectors of the first reference point and the second reference point, which are located on the same horizontal line; and/or

determining a second motion vector difference component based on a difference between the motion vectors of the third reference point and the fourth reference point, which are located on the same vertical line.

There is a first horizontal distance between the first control point and the second control point, and there is a first vertical distance between the third control point and the fourth control point. It should be noted that "and/or" in the decision in step S710 specifically indicates the following three different technical solutions:

Solution 1: A method for determining a motion vector difference of an affine image block based on three control points includes: determining a first motion vector difference component based on a difference between the motion vectors of the first control point and the second control point, which are located on the same horizontal line, and determining a second motion vector difference component based on a difference between the motion vectors of the third reference point and the fourth reference point, which are located on the same vertical line.

Solution 2: A method for determining a motion vector difference of an affine image block based on two control points in the horizontal direction includes: determining a first motion vector difference component based on a difference between the motion vectors of the first control point and the second control point, which are located on the same horizontal line.

Solution 3: A method for determining a motion vector difference of an affine image block based on two control points in the vertical direction includes: determining a second motion vector difference component based on a difference between the motion vectors of the third control point and the fourth control point, which are located on the same vertical line.

It should be understood that in this embodiment of the present invention, the first motion vector difference component is the horizontal motion vector difference component, and the second motion vector difference component is the vertical motion vector difference component. In Solution 1, the first motion vector difference component and the second motion vector difference component together reflect the motion vector difference of the affine image block. However, in Solution 2 and Solution 3, the first motion vector difference component and the second motion vector difference component independently reflect the motion vector difference of the affine image block. In other words, in Solution 1, the motion vector difference of the affine image block includes a first motion vector difference component and a second motion vector difference component; in solution 2, the motion vector difference of the affine image block includes the first motion vector difference component; and in solution 3, the motion vector difference of the affine image block includes a second motion vector difference component. It should also be understood that in this embodiment of the present invention, the first, second, third and fourth are only used to distinguish between pixels and should not constitute any limitation on the protection scope of the present invention. For example, the first reference point may also be referred to as the second reference point, and the second reference point may be referred to as the first reference point.

Further, in this embodiment of the present invention, the first reference point and the third reference point are the same pixels. More specifically, three control points are selected and one of the three control points is on the same horizontal line as one of the other two control points, and is located on the same vertical line as the last control point.

Perhaps, in this embodiment of the present invention, the first control point, the second control point, the third control point and the fourth control point are the vertices of the affine block. The four vertices in an affine block are assumed to be respectively labeled top left, bottom left, top right, and bottom right.

In solution 1, any three vertices in two orthogonal directions can be chosen to determine the difference in the motion vector of the affine block. For example, the top left vertex, bottom left vertex, and top right vertex may be selected, or the top left vertex, top right vertex, and bottom right vertex may be selected.

In solution 2, any two vertices located on the same horizontal line can be selected to determine the difference in the motion vector of the affine block. For example, the top left vertex and top right vertex may be selected, or the bottom left vertex and bottom right vertex may be selected.

In solution 3, any two vertices located on the same vertical line can be selected to determine the difference in the motion vector of the affine block. For example, the top left vertex and bottom left vertex may be selected, or the top right vertex and bottom right vertex may be selected. This is not limited in the present invention.

Possibly, in this embodiment of the present invention, the difference between the components that are the motion vectors of the first reference point and the second reference point, and which are in one of two directions (horizontal direction and vertical direction, or referred to as the x-direction and y-direction), may be defined as the first component of the difference of the motion vector of the affine block and the difference between the components that are the motion vectors of the third control point and the fourth control point, and in either of the two directions can be determined as the second component of the difference of the motion vector of the affine block. The method for determining the motion vector difference is applied to solutions 1, 2, and 3. Alternatively, according to the actual needs of coding complexity and coding performance, the values between two differences between the motion vector components of the first reference point and the second reference point in two directions may be is selected and determined as the first component, and a value between two differences between the motion vector components of the third reference point and the fourth reference point in two directions may be selected and determined as the second component. This is not limited in the present invention.

The motion vector of each from the first waypoint to the fourth waypoint typically includes two components, ie, a horizontal component and a vertical component. Thus, in the preceding steps, determining the first motion vector difference component based on the difference between the motion vectors of the first reference point and the second reference point, which are located on the same horizontal line, and determining the second motion vector difference component based on the difference between the motion vectors of the third reference point and the fourth control point, which are located on the same vertical line, can be concretely implemented as follows:

determining a first horizontal difference component and a first vertical difference component between the motion vectors of the first reference point and the second reference point, and determining more than one first horizontal difference component and the first vertical difference component as the first motion vector difference component; and

respectively, determining the second horizontal difference component and the second vertical difference component between the motion vectors of the third reference point and the fourth reference point, and determining more than one second horizontal difference component and the second vertical difference component as the second motion vector difference component.

и

используют для представления горизонтальной составляющей и вертикальной составляющей вектора движения пикселя на позиции

, разность вектора движения аффинного блока может быть определена в соответствии с формулами (13) и (14):More specifically, if

and

is used to represent the horizontal component and the vertical component of the motion vector of a pixel at a position

, the difference of the motion vector of the affine block can be determined in accordance with formulas (13) and (14):

(13)

(thirteen)

(14)

(fourteen)

представляет собой максимальное значение,

представляет собой горизонтальную составляющая разности вектора движения аффинного блока,

представляет собой вертикальную составляющая разности вектора движения аффинного блока, W представляет собой первое горизонтальное расстояние между первой контрольной точкой и второй контрольной точкой, и Н представляет собой первое вертикальное расстояние между третьей контрольной точкой и четвертой контрольной точкой. Со ссылкой на фиг.8, формула (13) выражает, что вектор движения горизонтальной разности

является максимальной дельтой компонентов векторов движения двух контрольных точек в одном и том же горизонтальном направлении, другими словами, абсолютное значение разности вектора движения является самым большим. Например, вершины 0 и 1 или вершины 2 и 3 могут быть выбраны в качестве контрольных точек, чтобы вычислить

. Формула (14) выражает, что вектор движения вертикальной разности

является максимальной дельтой компонентов векторов движения двух контрольных точек в одном и том же вертикальном направлении, другими словами, абсолютное значение разности вектора движения является самым большим. Например, вершины 0 и 2 или вершины 1 и 3 могут быть выбраны в качестве контрольных точек, чтобы вычислить

. Это может быть понятно, что формулы (13) и (14) применимы к любому из решений 1, 2 и 3. Детали представлены следующим образом:

represents the maximum value,

represents the horizontal component of the difference of the motion vector of the affine block,

is the vertical difference component of the motion vector of the affine block, W is the first horizontal distance between the first control point and the second control point, and H is the first vertical distance between the third control point and the fourth control point. Referring to FIG. 8, formula (13) expresses that the motion vector of the horizontal difference

is the maximum delta of the motion vector components of the two control points in the same horizontal direction, in other words, the absolute value of the motion vector difference is the largest. For example, vertices 0 and 1 or vertices 2 and 3 can be chosen as control points to calculate

. Formula (14) expresses that the motion vector of the vertical difference

is the maximum delta of the motion vector components of the two control points in the same vertical direction, in other words, the absolute value of the motion vector difference is the largest. For example, vertices 0 and 2 or vertices 1 and 3 can be chosen as control points to calculate

. It can be understood that formulas (13) and (14) apply to any of solutions 1, 2 and 3. Details are presented as follows:

In Solution 1, a motion vector difference of an affine image block is determined based on three control points, and the motion vector difference includes a first component and a second component. Thus, in solution 1, the motion vector difference of the affine image block is obtained by using a combination of formula (13) and formula (14).

In Solution 2, the motion vector difference of the affine image block is determined based on two control points in the horizontal direction, and the motion vector difference only includes the first component. Thus, in solution 2, the motion vector difference of the affine image block is obtained using formula (13). Fig. 8 is used as an example. In most cases, the motion vector of a vertex in an affine block is known. In this case, the motion vector difference can be directly computed based on the motion vector of the vertex in the affine block.

и

движения вершин в аффинной блоке на фиг. 8 известны, то:If the vectors

and

vertex movements in the affine block in Fig. 8 are known, then:

(15)

(fifteen)

и

движения вершин в аффинной блоке на фиг. 8 известны, то:If the vectors

and

vertex movements in the affine block in Fig. 8 are known, then:

(16)

(sixteen)

A combination of other known motion vectors may also be selected. This is not limited in the present invention.

These control point motion vectors used in the above solution can be obtained based on an affine transformation model. For example, when a 4-parameter affine transformation model is used, and (0,0) at the top left and (W, 0) at the top right in an affine image block are used as control points, the model formula is as follows:

(17)

(17)

и

, соответственно, являются векторами движения контрольных точек на верхнем левом углу и правом верхнем углу. Если координаты

и

контрольных точек в нижнем левом углу и нижнем правом углу подставить в формулу, то может быть получено

и

.

and

, respectively, are the motion vectors of the top left and top right control points. If the coordinates

and

control points in the lower left corner and lower right corner are substituted into the formula, then can be obtained

and

.

In Solution 3, the motion vector difference of the affine image block is determined based on two control points in the vertical direction, and the motion vector difference includes only the second component. Thus, in solution 3, the motion vector difference of the affine image block is obtained using formula (14). Fig. 8 is used as an example. In most cases, the motion vector of a vertex in an affine block is known. In this case, the motion vector difference can be directly computed based on the motion vector of the vertex in the affine block.

и

движения вершин в аффинной блоке на фиг. 8 известны, то:If the vectors

and

vertex movements in the affine block in Fig. 8 are known, then:

(18)

(eighteen)

и

движения вершин в аффинной блоке на фиг. 8 известны, то:If the vectors

and

vertex movements in the affine block in Fig. 8 are known, then:

(19)

(nineteen)

A combination of other known motion vectors may also be selected. This is not limited in the present invention.

These control point motion vectors used in the above solution can be obtained based on the affine transformation model. For example, when a 4-parameter affine transformation model is used and (0,0) at the top left and (0, H) at the bottom left are used as control points, the model formula is:

(20)

(20)

и

, соответственно, являются векторами движения контрольных точек на верхнем левом углу и в нижнем левом углу. Если координаты

и

контрольных точек на верхний правом углу и нижнем правом углу подставляют в формулу, то может быть получено

и

.

and

, respectively, are the motion vectors of the top left and bottom left control points. If the coordinates

and

control points on the upper right corner and lower right corner are substituted into the formula, then can be obtained

and

.

Perhaps, in this embodiment of the present invention, the motion vector difference is obtained based on the affine model, that is, the motion vector difference is obtained by determining the affine transformation parameter of a pixel in the affine image block. It should be noted that pixels in an affine image block have the same affine transformation parameter.

Accordingly, the above step of determining the first horizontal difference component and the first vertical difference component between the motion vectors of the first reference point and the second reference point is specifically: determining the first horizontal difference component and the first vertical difference component based on the affine transformation parameter and the first horizontal distance; and the preceding step of determining the second horizontal difference component and the second vertical difference component between the motion vectors of the third reference point and the fourth reference point is: determining the second horizontal difference component and the second vertical difference component based on the affine transform parameter and the first vertical distance.

It should be understood that the sequence numbers of the mentioned processes do not indicate the sequence of execution. The sequence of execution of the processes should be determined based on the functions and internal logic of the processes, and should not represent any limitation in the processes of implementing this embodiment of the present invention.

Possibly, in this embodiment of the present invention, the affine transform model may be a prior art 6-parameter affine transform model, or may be a prior art 4-parameter or 2-parameter affine transform model. The 6-parameter affine transformation model is mainly applied to the scenario in Solution 1 based on three control points, and the 4-parameter affine transformation model is mainly applied to the scenarios in Solution 2 and Solution 3 based on two control points. The following describes a method for obtaining a motion vector difference based on the affine transform parameters separately, using the 6-parameter affine transform model and the 4-parameter affine transform model as an example.

являются координатами пикселя

в текущем аффинном блоке изображения в текущем изображении,

являются координатами пикселя

, который находится в опорном изображении, и который соответствует пикселю

, и

являются параметрами аффинного преобразования. Если параметры аффинного преобразования известны, то может быть вычислена позиция

пикселя

в опорном изображении и, следовательно, может быть получен сигнал предсказания пикселя из опорного изображения.In the 6-parameter affine transformation model, it is assumed that

are pixel coordinates

in the current affine image block in the current image,

are pixel coordinates

, which is in the reference image, and which corresponds to the pixel

, and

are parameters of the affine transformation. If the parameters of the affine transformation are known, then the position can be calculated

pixel

in the reference image, and therefore a pixel prediction signal can be obtained from the reference image.

(21)

(21)

(22)

(22)

The affine transform parameters can usually be calculated based on the motion vectors of some pixels in the affine image block, and the horizontal component and the vertical component of the motion vector can be expressed using formulas (23) and (24), respectively:

(23)

(23)

(24)

(24)

Formula (21) is substituted into formula (23), and formula (22) is substituted into formula (24) to obtain the horizontal component and the vertical component of the motion vector of the pixel with coordinates (x, y), where the horizontal component and the vertical component are expressed by the formulas ( 25) and (26), respectively:

(25)

(25)

(26)

(26)

Formula (25) is substituted into formula (13), and formula (26) is substituted into formula (14). Thus, formulas (13) and (14) are transformed into formulas (27) and (28), respectively:

(27)

(27)

(28)

(28)

,

,

и

аффинного преобразованием.The first component and the second component of the difference of the motion vector of the affine block can be determined based on the distance between the control points and by specifying the parameters

,

,

and

affine transformation.

являются координатами пикселя

в аффинном блоке изображения в текущем изображении,

являются координатами пикселя

, который находится в опорном изображении, и который соответствует пикселю

, и

являются параметрами аффинного преобразования. Если параметры аффинного преобразования известны, то может быть вычислена позиция

пикселя

в опорном изображении и, следовательно, может быть получен сигнал предсказания для пикселя из опорного изображения:In the 4-parameter affine transformation model, it is assumed that

are pixel coordinates

in an affine image block in the current image,

are pixel coordinates

, which is in the reference image, and which corresponds to the pixel

, and

are parameters of the affine transformation. If the parameters of the affine transformation are known, then the position can be calculated

pixel

in the reference image, and therefore a prediction signal for a pixel from the reference image can be obtained:

(29)

(29)

(30)

(thirty)

Formula (29) corresponds to the scenario in solution 2, in which the motion vector difference of an affine image block is determined based on two control points in the horizontal direction. Formula (30) corresponds to the scenario in solution 3, in which the motion vector difference of an affine image block is determined based on two control points in the vertical direction.

The affine transform parameters can usually be calculated based on the motion vectors of some pixels in the image block. For the horizontal component and the vertical component of the motion vector, reference is made to formulas (23) and (24). Formulas (29) and (30) are substituted into formulas (23) and (24), respectively, to obtain the horizontal component and vertical component of the pixel motion vector with coordinates (x, y), where the horizontal component and vertical component are expressed by formulas (31) and (32) respectively:

(31)

(31)

(32)

(32)

Formulas (31) and (32) are substituted into formulas (13) and (14). Thus, formulas (13) and (14) are transformed into formulas (33) and (34), respectively:

(33)

(33)

(34)

(34)

The first component or second component of the motion vector difference of the affine block in the scenario in Decision 2 or Decision 3 may be determined based on the distance between control points and by determining the parameters a and b of the affine transformation.

It should be understood that the method for determining the motion vector difference using the 2-parameter affine transform model is essentially the same as the method for determining the motion vector difference used for the 6-parameter affine transform model or the 4-parameter affine transform model. For brevity, the details are not described here again.

Perhaps, in this embodiment of the present invention, in the decoding process, the affine transform parameters can be obtained by performing iterative parameter calculation. For example, the parameter a is incremented by 1 to determine if the model based on which the motion-compensated prediction signal is derived is optimal. Alternatively, the affine transformation parameters may be obtained by subtraction based on the affine transformation parameter of an adjacent affine block. However, this is not limited in the present invention.

Perhaps, in this embodiment of the present invention, the motion vector of the first reference point, the motion vector of the second reference point, the motion vector of the third reference point, and the motion vector of the fourth reference point can be determined to determine the motion vector difference of the affine image block.

Accordingly, in the scenario in Solution 1, the difference between the horizontal motion vector component of the first reference point and the horizontal motion vector component of the second reference point is determined as the first horizontal component of the difference; the difference between the vertical component of the motion vector of the first reference point and the vertical component of the motion vector of the second reference point is determined as the first vertical component of the difference; the difference between the horizontal component of the motion vector of the third reference point and the horizontal component of the motion vector of the fourth reference point is determined as the second horizontal component of the difference; and a difference between the vertical motion vector component of the third reference point and the vertical motion vector component of the fourth reference point are determined as the second vertical difference component.

In the scenario in Solution 2, the difference between the horizontal motion vector component of the first control point and the horizontal motion vector component of the second control point is determined as the first horizontal component of the difference; and the difference between the vertical component of the motion vector of the first reference point and the vertical component of the motion vector of the second reference point is determined as the first vertical component of the difference.

In the scenario in Solution 3, the difference between the horizontal motion vector component of the third reference point and the horizontal motion vector component of the fourth reference point is determined as the second horizontal component of the difference; and a difference between the vertical motion vector component of the third reference point and the vertical motion vector component of the fourth reference point are determined as the second vertical difference component.

In particular, the motion vectors of the control points can be directly determined, and the first component and the second component of the difference of the motion vectors of the affine block can be directly obtained by calculating the difference between the components of the motion vectors.

расположены на одной горизонтальной линии, и две контрольные точки C и D на расстоянии

расположены на одной вертикальной линии. Если векторы движения контрольных точек A, B, C и D известны, то первый компонент и второй компонент разности вектора движения аффинного блока могут быть определены в соответствии с формулами (35) и (36):For example, FIG. 9 is a diagram of an affine image block and control points in accordance with an embodiment of the present invention. Two control points A and B in the distance

are on the same horizontal line, and the two control points C and D are at a distance

located on the same vertical line. If the motion vectors of control points A, B, C and D are known, then the first component and the second component of the difference of the motion vector of the affine block can be determined in accordance with formulas (35) and (36):

(35)

(35)

(36)

(36)

Possibly, in this embodiment of the present invention, reference points A and C may be the same pixels.

Possibly, in this embodiment of the present invention, control points A, B, C, and D may be affine block vertices. In this case, the distance between control points A and B is the width W of the affine block, and the distance between control points C and D is the height H of the affine block.

Perhaps, in this embodiment of the present invention, the first control point and the second control point are two adjacent pixels, and the third control point and the fourth control point are two adjacent pixels. In other words, the control points are not necessarily chosen from the four corner points, also referred to as vertices, in the affine image block, but can be points that are at any position in the affine image block and are jointly set on the encoder side and the decoder side, explicitly specified in bitstream using signaling, or obtained by estimation or duplication based on breakpoints in the surrounding affine image block. In this case, the definitions of the first horizontal difference component, the first vertical difference component, the second horizontal difference component, and the second vertical difference component, are specifically as follows:

A first pixel motion vector, a second pixel motion vector, and a third pixel motion vector are determined, where the first pixel, the second pixel, and the third pixel are pixels that do not overlap with each other; determining a second horizontal distance and a second vertical distance between the first pixel and the second pixel; determining a third horizontal distance and a third vertical distance between the first pixel and the third pixel; the first horizontal difference component, the first vertical difference component, the second horizontal difference component, and the second vertical difference component are determined based on the motion vector of the first pixel, the motion vector of the second pixel, the motion vector of the third pixel, the second horizontal distance, the second vertical distance, the third horizontal distance, and the third vertical distance. The above description is based on solution 1 of three control points as an example. The difference between solution 2 or solution 3 based on two control points and solution 1 based on three control points is: use only the first difference component or the second difference component in solution 1 as the difference of the motion vector of the affine image block based on two control points.

,

и

, соответственно. Расстояние между пикселем A и пикселем В в горизонтальном направлении равно

и расстояние между пикселем A и пикселем В в вертикальном направлении равно

. Расстояние между пикселем A и пикселем С в горизонтальном направлении равно

, и расстояние между пикселем A и пикселем С в вертикальном направлении равно

. Расстояние между пикселем B и пикселем С в горизонтальном направлении равно

, и расстояние между пикселем B и пикселем С в вертикальном направлении равно

. Предполагают, что разности между горизонтальными составляющими векторов движения двух пикселей, которые являются смежными в горизонтальном направлении равны

и

, соответственно, и разности между вертикальными составляющими векторов движения двух пикселей, которые являются смежными в вертикальном направлении равны

и

, соответственно. Так как вектор движения в блоке изображения изменяется линейно,

,

,

и

могут быть определены путем определения векторов движения пикселей А, В и С, и может быть определена разность вектора движения между соседними пикселями в блоке изображения. В частности, разность вектора движения может быть определена по формулам (37) до (40):Specifically, A, B, and C are assumed to be in an image block of any three pixels that do not overlap with each other, the positions of the three pixels are shown in FIG. 10 and the motion vectors of the three pixels are

,

and

, respectively. The distance between pixel A and pixel B in the horizontal direction is

and the distance between pixel A and pixel B in the vertical direction is

. The distance between pixel A and pixel C in the horizontal direction is

, and the distance between pixel A and pixel C in the vertical direction is

. The distance between pixel B and pixel C in the horizontal direction is

, and the distance between pixel B and pixel C in the vertical direction is

. It is assumed that the differences between the horizontal components of the motion vectors of two pixels that are adjacent in the horizontal direction are

and

, respectively, and the difference between the vertical components of the motion vectors of two pixels that are adjacent in the vertical direction are

and

, respectively. Since the motion vector in the image block changes linearly,

,

,

and

can be determined by determining the motion vectors of pixels A, B, and C, and the motion vector difference between adjacent pixels in the image block can be determined. In particular, the motion vector difference can be determined by formulas (37) to (40):

(37)

(37)

(38)

(38)

(39)

(39)

(40)

(40)

Alternatively, the motion vector difference can be determined according to formulas (37) and (38) and formulas (41) and (42):

(41)

(41)

(42)

(42)

Possibly, in this embodiment of the present invention, the decoding device may obtain the motion vectors of all control points through motion estimation and search, may obtain the motion vectors of all control points based on the surrounding image block, or may calculate the motion vectors of all control points based on an affine transformation model. . Alternatively, the decoding device may obtain the motion vectors of some control points through motion estimation and searching, and obtain the motion vectors of other control points based on the neighboring image block. Alternatively, the decoder may obtain the motion vectors of some control points using affine motion estimation and search, and calculate the motion vectors of other control points based on the affine transformation model. Alternatively, the decoder may derive the motion vectors of some control points based on an adjacent image block and calculate the motion vectors of other control points based on the affine transform model. However, this is not limited in the present invention.

Perhaps, in S720, the first set value can be determined as the motion vector precision of the affine image block; or, the motion vector accuracy of the affine image block may be determined based on a feature of the adjacent image block of the affine image block. An adjacent image block is an image block adjacent to an affine image block in space and/or time.

In particular, the value of the motion vector of the affine block may be an integer. In this case, the motion vector precision is considered as integral pixel precision, that is, the pixel precision is 1. Alternatively, the motion vector precision value cannot be an integer. In this case, the motion vector is taken as a sub-pixel precision including a precision such as 1/2, ¼ or 1/8. Fig. 11 shows a 4×4 block image where x is a pixel at an integer pixel position and the collected image has only pixels at integer positions; and O is a pixel at 1/2 position precision, and a pixel at 1/2 position precision is to be obtained by performing interpolation on a pixel at an integer pixel position. The pixel value at the other precision position must be obtained by performing additional interpolation on an integer precision pixel or a 1/2 precision pixel. If the motion vector precision of the current pixel is an integer, the current pixel is indicated at the X position in the reference image. If the motion vector precision of the current pixel is 1/2, then the current pixel is specified at position O in the reference image.

.The motion vector precision of an affine block is the highest motion vector precision for all pixels in the affine block. The motion vector precision of the affine block may be specified, for example, by an integer pixel precision or by a precision such as 1/2, 1/4, 1/8, or 1/16. Alternatively, the motion vector accuracy of the affine block may be determined based on a feature of the neighboring image block of the affine block. For example, if the surrounding image is relatively smooth, then the current affine block can also be predicted to be smooth, and a relatively high motion vector precision, such as 1/8 or 1/16 precision, can be selected; otherwise, select a relatively low motion vector precision, such as an integer pixel precision or 1/2 precision. The resulting accuracy is denoted as

.

Perhaps, as shown in Fig. 12, S730 specifically includes the following steps.

S1201. To proportionally adjust the first horizontal distance based on the proportion of the motion vector accuracy to the first motion vector difference component to obtain the length of the affine motion compensation image subblock in the horizontal direction; and determine whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction direction, adjust the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal motion compensation affine image subblock.

In particular, the ratio of the motion vector accuracy to the first motion vector difference component is the ratio of the motion vector accuracy to the first motion vector difference component. The product of the ratio and the first horizontal distance is the length of the affine sub-image compensation motion in the horizontal direction, and can also be referred to the initial length in the horizontal direction. Obviously, the initial length of the affine subblock of the horizontal motion compensation image may not be an integer in some cases. In this case, the initial length must be rounded off. To ensure predictable accuracy, rounding down is generally performed. More specifically, the decimal part is directly discarded. For example, the initial length value is 8.35 (that is, the length is 8.35 unit pixels), and the number 8 is directly used as the length of the horizontal motion compensation image affine subblock. Of course, the rounding principle can be alternatively used. More specifically, if the fractional part is less than 5, rounding down is performed; and, if the fractional part is greater than 5, rounding up is performed. For example, when the length is calculated as 8.86, then 9 is used as the length. To ensure performance, the minimum value of the initial length should not be less than the specified value, for example, not less than the specified value, such as 2, 4, 8 or 2 ^N . The predetermined value is usually set to 4. The initial length should also not be greater than the length of the affine image block in the horizontal direction. Whether the initial length in the horizontal direction is correct or the initial length is rounded in the horizontal direction depends on whether the length of the affine image block in the horizontal direction is divisible by the initial length. If the length of the affine image block in the horizontal direction is divided by the initial length, then this indicates that the affine image block can be divided into an integer number of sub-blocks in the horizontal direction based on the initial length in the horizontal direction or the rounded initial length in the horizontal direction. The correction method is generally a method in which the initial length in the horizontal direction or the rounded initial length in the horizontal direction is gradually increased/decreased. More specifically, if it is determined that the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine sub-image sub-image of the horizontal direction motion compensation, the length of the affine sub-image of the motion compensation in the horizontal direction is corrected, that is, increased/decreased, by one or several units of length, and then cycling through said decision until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal motion compensation. The length unit is typically one block of pixels. More specifically, the length of the affine motion compensation image sub-frame in the horizontal direction is corrected, i.e., increased/decreased by 1. Without loss of generality, in other words, the length of the affine motion compensation image sub-frame in the horizontal direction is corrected, i.e., increased/decreased by a fixed step size. The step size is an adjustment value, i.e., a basic block or amplitude, used to correct the length of the affine sub-frame of the motion compensation image in the horizontal direction in each cycle. The step size is typically the aforementioned one pixel block, or may be another value such as two pixel blocks. It is possible that the correction size/step value used in the above correction may change depending on the ratio of the original length of the horizontal motion compensation affine sub-image subblock. For example, when the initial length of the affine subblock of the horizontal motion compensation image is an odd number, an odd step size is selected. For example, an odd number of length units, such as one length unit, is used as the step size for adjustment. When the initial length of the affine subblock of the horizontal motion compensation image is an even number, an even step size is selected. For example, an even number of unit lengths is used as the step size.

In addition, whether the adjustment is an increase or decrease can be selected according to an agreement between the decoder side and the encoder side, can be explicitly indicated on the encoder side in the bitstream, and is selected on the decoder side based on the explicit indication, can be inferred using the value accuracy of the motion vector of the surrounding neighboring block, or may be determined based on the relation of the ratio between the length of the affine image block in the horizontal direction and the length of the affine image sub-block of the motion compensation image in the horizontal direction. For example, if the ratio is less than a threshold value, decrease is selected; otherwise, increase is selected.

To simplify the algorithm and ensure the efficiency of encoding and decoding, if it is determined that the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, the length of the affine sub-image sub-image of the compensation motion in the horizontal direction is corrected, that is incremented by one predetermined length unit, and then the preceding decision will not be cycled until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion compensation.

Further, in order to increase efficiency and reduce operating time, the length of the affine sub-image block of the motion compensation in the horizontal direction may be adjusted to the nearest value by which the length of the affine image block in the horizontal direction is divided. For example, the length of the affine image subblock of the horizontal direction motion compensation image is 7, while the length of the affine image block of the horizontal direction is 16. In this case, the length of the affine image subblock of the image compensation in the horizontal direction is increased by changes 1 to 8, with by which 16 is divided. Similarly, if the length of the horizontal direction motion compensation affine image subblock is 5, the length of the horizontal direction motion compensation affine image subblock is reduced by 1 to change to 4, by which 16 is divided. If the differences between the length of the horizontal direction motion compensation affine image block and two adjacent values by which the horizontal direction affine image block length is divided are equal, for example, the length of the horizontal direction motion compensation affine image block is 6, the length of the affine image block in the horizontal direction is 16, and two values that are adjacent with a length of 6 and by which 16 is divided are 4 and 8, respectively, the correction can be selected according to the requirements of the encoder side and the decoder side. For example, if a low latency is required, an adjustment, i.e. an increase, may be chosen. If high quality is required, choose an adjustment, that is, a reduction. This is not limited in the present invention.

, и Wmax является возможно максимальной длиной аффинного блока изображения в горизонтальном направлении. Определяют интервал, в котором находится длина аффинного субблока изображения компенсации движения в горизонтальном направлении, и значение нижнего предела или значение верхнего предела интервала, в рамках которого длина аффинного субблока изображения компенсации движения в горизонтальном направлении, используют как окончательно определенную длину аффинного субблока изображения компенсации движения в горизонтальном направлении. Для правила получения значения нижнего предела или значения верхнего предела, относится к соглашению между стороной кодера и стороной декодера. Верхнее предельное значение или нижнее предельное значение приближенного интервала может быть явно указано, используя поток битов, или может быть выбрано посредством округления. С помощью приведенных выше таблиц, окончательная длина аффинного субблока изображения компенсации движения в горизонтальном направлении может быть определена с помощью таблицы поиска на основе начальной длины аффинного субблока изображения компенсации движения в горизонтальном направлении. Таким образом, этап корректировки длины аффинного субблока изображения компенсации движения в горизонтальном направлении на этапе S1201 можно заменить, сложность системы снижается и время вычисления уменьшается.Further, in step S1201, when it is determined that the length of the affine picture block in the horizontal direction is not an integer multiple of the length of the affine picture sub-picture compensation in the horizontal direction, in order to further simplify the algorithm, the same table can be separately set at the encoder side, and side of the decoder, and the possible maximum length of the affine image block in the horizontal direction is divided into a plurality of intervals, using the allowed minimum divisor as the initial value, for example, 2 ^N . The duration of the interval is 2 ^N+i , where

, and Wmax is the possible maximum length of an affine image block in the horizontal direction. The interval in which the length of the affine sub-frame of the motion compensation image in the horizontal direction is located is determined, and the value of the lower limit or the value of the upper limit of the interval within which the length of the affine sub-frame of the motion compensation image in the horizontal direction is used as the finally determined length of the affine sub-frame of the motion compensation image in horizontal direction. For a rule for obtaining a lower limit value or an upper limit value, refers to an agreement between the encoder side and the decoder side. The upper limit value or the lower limit value of the approximate interval may be explicitly indicated using the bitstream, or may be selected by rounding. With the above tables, the final length of the horizontal direction motion compensation affine image subblock can be determined with a lookup table based on the initial length of the horizontal direction motion compensation affine image subblock. Thus, the step of adjusting the length of the affine sub-frame of the horizontal direction motion compensation image in step S1201 can be replaced, the complexity of the system is reduced, and the calculation time is reduced.

It is possible to further reduce the computational complexity in the decision, in step S1201, but not only if the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine sub-image sub-image compensation of movement in the horizontal direction, it is determined whether the length of the affine image sub-image sub-image of the compensation movement in the horizontal direction is an integer multiple of the given length S, which facilitates the calculation. More specifically, whether it is to be determined whether the length of the affine subblock of the horizontal motion compensation image meets the following conditions:

(1) The length of the affine image block in the horizontal direction is an integer multiple of the length of the affine sub-image block of the horizontal motion compensation; and (2) the length of the affine subblock of the horizontal motion compensation image is an integer multiple of the predetermined length S.

If the length of the affine motion compensation image subblock in the horizontal direction does not meet the two conditions, the length of the affine motion compensation image subblock in the horizontal direction is adjusted so that the length of the affine subblock of the motion compensation image in the horizontal direction satisfies condition (1) and condition (2). More specifically, the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine image sub-image compensation in the horizontal direction, and also to correct the length of the affine image sub-image compensation in the horizontal direction is an integer multiple of the predetermined length. The given length S is typically 4, or may be a value such as 8, 16, or 32. In other words, S = 2 ^N , where N is zero or a positive integer. This solution is applicable to an affine image block in any form. When an affine image block is partitioned according to the Quad Tree rule, provided that any of the above conditions defining (1) and (2) is satisfied, another condition may be satisfied. Thus, only one of the defining conditions must be activated/applied. The sequence for determining whether conditions (1) and (2) are met and the sequence for adjusting the length of the affine motion compensation image sub-block in the horizontal direction based on the determination result may be randomly interchanged. For distinction, the adjustment made during the first time may be referred to as the first adjustment, and the adjustment made at the second time (if necessary) may be referred to as the re-adjustment.

Alternatively, a length amplitude limiting operation, which is the affine sub-block of the horizontal direction motion compensation image, and as defined above, should be performed. More specifically, a comparison method is used to ensure that M is within the range from the lower limit Tm to the upper limit W. Tm is a predetermined value and is generally 4, and W is the length of the affine image block in the horizontal direction.

S1203. To proportionally adjust the first vertical distance based on the proportion of the motion vector accuracy to the second motion vector difference component to obtain the length of the affine motion compensation image subblock in the vertical direction; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation and, if the length of the affine image block in the vertical direction is not an integer, a multiple of the length of the affine image sub-image subblock in the vertical direction , adjust the length of the affine image subblock of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the adjusted length of the affine image subblock of the vertical motion compensation image.

In particular, the ratio of the motion vector accuracy to the second motion vector difference component is the ratio of the motion vector accuracy to the second motion vector difference component. The product of the ratio and the first vertical distance is the length of the vertical motion compensation affine sub-image subblock, and can be referred to the initial length in the vertical direction. It may be appreciated that the initial length of the vertical motion compensation affine subblock image may not be an integer in some cases. In this case, the initial length must be rounded off. To ensure accuracy, round down as a rule. More specifically, the decimal part is directly discarded. For example, the value of the initial length is 8.35 (that is, the length is 8.35 unit pixels), and 8 is directly used as the length of the vertical motion compensation image affine subblock. Of course, the principle of rounding may alternatively be used. More specifically, if the fractional part is less than 5, rounding down is performed; and, if the fractional part is greater than 5, then rounding up is performed. For example, when the length is calculated as 8.86, 9 is used as the length. To ensure performance, the minimum value of the initial length should not be less than the specified value, for example, not less than the specified value, such as 2, 4, 8 or 2 ^N . The set value is usually set to 4. The initial length should also not be greater than the length of the affine image block in the vertical direction. Whether the initial length in the vertical direction is correct or the rounded initial length in the vertical direction depends on whether the length of the affine image block in the vertical direction is divisible by the initial length. If the length of the affine image block in the vertical direction is divided by the initial length, this indicates whether the affine image block can be divided into an integer number of sub-blocks in the vertical direction based on the initial length in the vertical direction or the rounded initial length in the vertical direction. The correction method is generally a method in which the initial length in the vertical direction or the rounded initial length in the vertical direction is gradually increased/decreased. More specifically, if it is determined that the length of the vertical direction affine image block is not an integer multiple of the length of the vertical direction motion compensation affine image subblock, the length of the vertical direction motion compensation image affine subblock, that is, increase/decrease, is corrected by one or several unit lengths, and then the foregoing decision is cycled until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. The length unit is typically one block of pixels. More specifically, the length of the vertical direction motion compensation image affine subblock is corrected, that is, increased/decreased by 1. Without loss of generality, in other words, the length of the vertical motion compensation image affine subblock is corrected, that is, increased/decreased by a fixed step size. The step size is a correction value, that is, a base block or an amplitude used to correct the length of the affine sub-block of the vertical motion compensation image in each cycle. The step size is typically one pixel block, or may be another value such as two pixel blocks. It is possible that the step size/correction step used in the above correction may vary depending on the ratio of the initial length of the vertical motion compensation affine sub-image sub-block. For example, when the initial length of the affine subblock of the vertical motion compensation image is an odd number, an odd step size is selected. For example, an odd number of length units, such as one length unit, is used as the step size for adjustment. When the initial length of the affine subblock of the vertical motion compensation image is an even number, an even step size is selected. For example, an even number of unit lengths is used as the step size.

Additionally, whether the adjustment is an increase or decrease can be selected according to an agreement between the decoder side and the encoder side, can be explicitly indicated on the encoder side in the bitstream, and is selected on the decoder side based on the explicit indication, can be inferred using the precision of the vector motion of the surrounding neighboring block, or may be determined based on the ratio of the ratio between the length of the affine image block in the vertical direction and the length of the affine sub-image block of the compensation movement in the vertical direction. For example, if the ratio is less than the threshold, decrease is selected; otherwise, select magnification.

To simplify the algorithm and ensure the efficiency of encoding and decoding, if it is determined that the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine sub-image sub-block of the vertical direction motion compensation image, the length of the affine sub-image sub-block of the image compensation in the vertical direction is corrected, that is increase by one predetermined length unit and then perform said decision cyclically until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation.

Further, in order to increase efficiency and reduce operation time, the length of the affine sub-image block of the vertical direction motion compensation can be brought to the nearest value by which the length of the affine image block in the vertical direction is divided. For example, the length of the vertical direction motion compensation affine image subblock is 7, while the length of the vertical direction affine image block is 16. In this case, the length of the vertical direction compensation affine image subblock is increased by 1 to change to 8, with which 16 is divisible. Similarly, if the length of the vertical motion compensation affine sub-image subframe is 5, the length of the vertical motion compensation affine sub-image subframe is reduced by 1 to change to 4, by which 16 is divided. If the difference between the length of the vertical direction motion compensation affine image subblock and two adjacent values by which the vertical direction affine image block length is divided are equal, for example, the length of the vertical direction motion compensation affine image subblock is 6, the length of the affine image block in vertical direction is 16, and two values that are adjacent with a length of 6 and by which 16 is divided equal to 4 and 8, respectively, the correction can be selected according to the requirements of the encoder side and the decoder side. For example, if a low latency is required, an adjustment, i.e. an increase, may be chosen. If high quality is required, then an adjustment is chosen, that is, a reduction. This is not limited in the present invention.

, а Wmax является возможной максимальной длиной аффинного блока изображения в вертикальном направлении. Определяют интервал, в рамках которого находится длина аффинного субблока изображения компенсации движения в вертикальном направлении, и значение нижнего предела или значение верхнего предела интервала, в рамках которого находится длина аффинного субблока изображения компенсации движения в вертикальном направлении, используются как окончательная определенная длина аффинного субблока изображения компенсации движения в вертикальном направлении. Для правила получения значения нижнего предела или значения верхнего предела сделана ссылка на соглашение между стороной кодера и стороной декодера. Верхнее предельное значение или нижнее предельное значение приближенного интервала может быть явно указано, используя поток битов, или может быть выбрано посредством округления. С помощью приведенных выше таблиц, окончательная длина аффинного субблока изображения компенсации движения в вертикальном направлении может быть определена с помощью таблицы поиска на основе начальной длины аффинного субблока изображения компенсации движения в вертикальном направлении. Таким образом, этап корректировки длины аффинного субблока изображения компенсации движения в вертикальном направлении на этапе S1203 можно заменить, сложность системы снижается, и время вычисления уменьшается.Further, in step S1203, when it is determined that the affine image block length in the vertical direction is not an integer multiple of the length of the vertical motion compensation affine sub-image block, to further simplify the algorithm, the same table may be separately set on the encoder side and the decoder, and the possible maximum length of the affine image block in the vertical direction is divided into a plurality of intervals, using the allowable minimum divisor as the initial value, for example, 2 ^N . The duration of the interval is 2 ^N+i , where

, and Wmax is the possible maximum length of an affine image block in the vertical direction. The interval within which the length of the vertical motion compensation affine image subblock is located is determined, and the lower limit value or the upper limit value of the interval within which the vertical motion compensation image affine subblock length is located is used as the final determined length of the compensation image affine subblock. movement in the vertical direction. For the rule for obtaining the lower limit value or the upper limit value, reference is made to the agreement between the encoder side and the decoder side. The upper limit value or the lower limit value of the approximate interval may be explicitly indicated using the bitstream, or may be selected by rounding. With the above tables, the final length of the affine sub-image sub-frame of the vertical motion compensation image can be determined with a lookup table based on the initial length of the affine sub-frame of the vertical motion compensation image. Thus, the step of adjusting the length of the affine sub-block of the vertical motion compensation image in step S1203 can be replaced, the complexity of the system is reduced, and the calculation time is reduced.

It is possible to further reduce the computational complexity in the decision in step S1203 not only if the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the vertical motion compensation image, but also if the length of the affine image subblock of the image compensation in the vertical direction be an integer multiple of the given length S, which makes it easier to calculate. More specifically, it is necessary to determine whether the length of the affine sub-block of the vertical motion compensation image satisfies the following conditions:

(1) The length of the affine image block in the vertical direction is an integer multiple of the length of the affine sub-image block of the vertical motion compensation; and (2) the length of the affine subblock of the vertical motion compensation image is an integer multiple of the predetermined length S.

If the length of the vertical direction motion compensation image affine subblock does not meet the two conditions, the length of the vertical direction motion compensation image affine subblock is adjusted so that the length of the vertical direction motion compensation image affine subblock satisfies condition (1) and condition (2). More specifically, the affine image block length in the vertical direction is an integer multiple of the corrected length of the vertical motion compensation affine image sub-image, and the corrected length of the vertical motion compensation affine image sub-image is an integer multiple of the predetermined length. The given length S is typically 4, or may be a value such as 8, 16, or 32. In other words, S = 2 ⁿ , where n is zero or a positive integer. This solution is applicable to an affine image block in any form. When an affine image block is divided according to a Quadtree rule, provided that any of the above definition conditions (1) and (2) is satisfied, another condition may be satisfied. Thus, only one of the defining conditions must be activated/applied. The sequence for determining the fulfillment of the condition (1) and (2) and the sequence for correcting the length of the affine sub-block of the motion compensation image in the horizontal direction based on the determination result may be interchanged randomly. For distinction, the adjustment made during the first time may be referred to as the first adjustment, and the adjustment made at the second time (if necessary) may be referred to as the re-adjustment.

The above steps S1201 and S1203 are applicable to the scenario in decision 1 based on the three checkpoints. In the scenario, based on the three control points, the lengths of the motion compensation affine sub-image in the horizontal direction and the vertical direction are calculated separately. For the scenarios in solution 2 and solution 3 based on two breakpoints, only step S1201 or step S1203 can be performed to independently calculate the horizontal direction motion compensation affine sub-image sub-frame length or the vertical direction motion compensation affine image sub-frame length. Then, the length of the horizontal/vertical direction motion compensation image affine subblock or the adjusted length of the horizontal/vertical direction motion compensation image affine subblock is used as the initial length of the motion compensation image affine subblock in the other direction, i.e., the vertical/horizontal direction, and separately fulfill the above definitions, that is:

determining whether the length of the affine image block in the vertical/horizontal direction is an integer multiple of the length of the affine sub-image block of the vertical/horizontal direction motion compensation; or

determining whether the length of the affine image block in the vertical/horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the vertical/horizontal direction motion compensation, and determining whether the length of the affine sub-image sub-block of the motion compensation in the vertical/horizontal direction is an integer multiple of the predetermined length S.

In the scenario of decision 2 and decision 3 based on two control points, according to the requirement, only the length of the affine sub-image sub-frame of the motion compensation in the horizontal direction can be calculated in the solution with three control points, or only the length of the affine sub-image of the motion compensation in the vertical direction can be calculated; and another component size of the affine sub-block of the motion compensation image is output or calculated based on one component, the obtained size, that is, the length in the horizontal direction or the length in the vertical direction. Obviously, when the size of the affine image block is the shape of a regular 2Nx2N square, since the lengths of the affine image block in the horizontal direction and the vertical direction are the same, then only one of the lengths of the affine image block in the horizontal direction and the vertical direction will need to be determined.

The operation of limiting the amplitude along the length, which is the length of the affine subblock of the vertical motion compensation image, may need to be performed, which is determined in the above manner. More specifically, a comparison method is used to ensure that M is within the range from the lower limit Tm to the upper limit W. Tm is a predetermined value and is generally equal to 4, and W is the length of the affine image block in the vertical direction.

и

вектора движения, как правило, сравнительно невелики, что свидетельствует о том, что точность вектора движения каждого пикселя чрезмерно высока, например, 1/32 или 1/64. Чрезмерно высокая точность вектора движения мало способствует производительности сжатия видео, и вызывает огромное количество вычислений. Следовательно, аффинный блок делят на множество аффинных-MC блоков на основании ранее полученной ожидаемой точности

вектора движения, чтобы гарантировать, что точность вектора движения каждого аффинного MC блока достигает

. Меньшее значение

и

не меньше

.In particular, the resulting differences

and

motion vectors are typically relatively small, indicating that the motion vector accuracy of each pixel is excessively high, such as 1/32 or 1/64. Excessively high motion vector precision contributes little to video compression performance, and causes a huge amount of computation. Therefore, the affine block is divided into a set of affine-MC blocks based on the previously obtained expected accuracy

motion vector to ensure that the motion vector accuracy of each affine MC block reaches

. lower value

and

not less

.

или обозначается как М, и высота равна

или обозначается как N.Perhaps, in this embodiment of the present invention, it is assumed that the horizontal distance between the first reference point and the second reference point is W, and the vertical distance between the third reference point and the fourth reference point is H, and the affine MC block width is assumed to be

or denoted as M, and the height is

or denoted as N.

In the scenario in Solution 1, determining the motion vector difference of the affine image block based on the three control points and obtaining the size of the affine sub-image sub-block of the motion compensation image can be realized by the following formula (43), that is, M and N are determined according to formula (43) or (44):

и

(43); или

and

(43); or

и

(44)

and

(44)

This is equivalent to:

(45)

(45)

In the scenario in Solution 2, determining the motion vector difference of the affine image block based on two control points in the horizontal direction and obtaining the size of the affine sub-image sub-block of the motion compensation image can be realized by the following formula (46), that is, M and N are determined according to the formula ( 46):

(46)

(46)

This is equivalent to:

(47)If a

(47)

In the scenario in Solution 3, determining the motion vector difference of the affine image block based on two control points in the vertical direction and obtaining the size of the affine sub-image sub-block of the motion compensation image can be implemented using the following formula (48), that is, M and N are determined according to the formula ( 48):

(48)

(48)

This is equivalent to:

(49)If a

(49)

After calculating M and N according to formulas (45), (47) and (49), to simplify the algorithm, the M and H operations are facilitated, namely, it is required to additionally determine whether an affine image block with a width W and a height H can be evenly divided by an integer number of affine sub-blocks of the motion compensation image. More specifically, it is necessary to determine whether W can be divisible by M and H is divisible by N. If W is indivisible by M and/or H is indivisible by N, the value by which W and/or H is indivisible, that is, M and /or N are adjusted so that W is divisible by M, H is divisible by N, and M and N satisfy Tm ≤ M ≤ W and Tn ≤ N ≤ H.

The correction method can be as follows:

(1) When W is indivisible by M, M is gradually decreased/increased until W is divisible by M. When H is indivisible by N, N is gradually decreased/increased until H is divisible by N. When M = N , or only M and N must be adjusted.

Perhaps the upper limit and lower limit of M and the upper limit and lower limit of N are limited, so that Tm ≤ M ≤ W, Tn ≤ N ≤ H, Tm and Tn are usually set to 4.

(2) Using a lookup table: in the constructed two-dimensional table, W and M are input, and output M that satisfies the condition; and H and N are inputs, and an N that satisfies the condition is output.

Perhaps, after calculating M and N by formulas (45), (47) and (49), M and N are further corrected so that W is divisible by M, H is divisible by N, and M and H are each a multiple of S and Tm ≤ M ≤ W and Tn ≤ N ≤ H are satisfied.

The correction method can be as follows:

(1) When W is indivisible by M or M is not a multiple of S, M is decremented until W is divisible by M and M is a multiple of S. If H is indivisible by N or N is not a multiple of S, then N is decremented until then, H will not be divisible by N and N is a multiple of S. S is usually set to 4. When M = N, only one of M and N needs to be adjusted.

Perhaps the upper limit and lower limit of M and the upper limit and lower limit of N are limited, so that Tm ≤ M ≤ W, Tn ≤ N ≤ H, Tm and Tn are usually set to 4.

(2) The lookup table method is used: in the constructed two-dimensional table, W and M are input, and output M that satisfies the condition; and H and N are input and output N that satisfies the condition.

In particular, the above solutions can be implemented using the following example code algorithm in Table 1.

That is, according to the video image decoding method in this embodiment of the present invention, the decoding apparatus determines the size of the motion compensation affine image sub-block based on the determined difference of the motion vector and the motion vector accuracy of the affine image block, the distance between control points, the size of the affine image block, so that an affine image block can be evenly divided into a plurality of affine motion compensation image sub-blocks with the same size. Thus, the method is implemented in hardware/software and also the complexity is reduced. It has been proved by experiments in the present invention that the complexity on the encoder side can be reduced by 50%, the complexity on the decoder side can be reduced by 70%, and 90% - 95% of the performance is maintained. Thus, the motion compensation prediction technology based on the affine model can be effectively applied in practice.

Perhaps, as shown in Fig. 13, S740 includes the following steps:

S1301. Determine the motion vector of each affine subblock of the motion compensation image in the affine subblocks of the motion compensation image.

S1303. Determine the motion compensation prediction signal of each affine motion compensation image subblock based on the motion vector of each affine motion compensation image subblock.

S1305. Decode each affine sub-block of the motion compensation image based on the motion-compensated prediction signal of each affine sub-block of the motion compensation image.

It should be understood that in this embodiment of the present invention, each affine MC block may include a plurality of pixels, and a pixel motion vector must be selected from each affine MC block as the motion vector of the affine MC block.

Possibly, in S1301, the motion vector of a pixel at the center position of the affine MC block can be selected as the motion vector of the affine MC block, the average of the motion vectors of all pixels in the affine MC block can be used as the motion vector of the affine MC block, or the motion vector of the top the left pixel in the affine MC block can be chosen as the motion vector of the affine MC block. However, this is not limited in the present invention.

Possibly, in this embodiment of the present invention, the border pixel signal in each affine subblock of the motion compensation image is filtered, and the border pixel is one or more rows of pixels on the border of each affine subblock of the motion compensation image.

Possibly, in this embodiment of the present invention, the boundary pixel signal includes a motion-compensated prediction signal and/or a reconstructed signal, and the reconstructed signal is the sum of the motion-compensated prediction signal and the reconstructed residual signal.

In particular, the block indicated in the diagram by a solid thick line is an affine block, the block indicated by a solid thin line in the diagram is an affine MC block, the block indicated by a dotted line in the diagram is a pixel on the border of an adjacent affine MC block, and the intersection point is a pixel. The area of the block, indicated by the dotted line in FIG. 11 includes two rows or two columns of pixels in adjacent affine MC blocks on respective boundaries of adjacent affine MC blocks, or may include one row or one column, three rows or three columns, or the like of pixels on respective boundaries of adjacent affine MC blocks. Since the motion vectors of adjacent affine MC blocks may be different, the prediction signals obtained from the reference picture are not contiguous in the reference picture. This causes discontinuity of the prediction signals at the boundaries of adjacent affine MC blocks, which further causes discontinuity in the residual signal, and affects the image encoding/residual decoding performance. Thus, the filtering of the prediction signal with motion compensation at the boundary of the affine MC block is considered.

The recovered signal is typically obtained by summing the motion-compensated prediction signal and the recovered residual signal. Lossy coding is usually performed on the residual signal. This results in a distortion of the reconstructed residual signal with respect to the original residual signal. Pixel distortion at the boundaries of adjacent affine MC blocks can be different. For example, a pixel value on the right side in the first affine MC block becomes larger due to distortion, and a pixel value on the left side in an affine MC block that is adjacent to and to the right of the first affine MC block becomes smaller due to distortion. This causes a break in the boundary values of the reconstructed pixels in the affine MS blocks, and distorts the subjective and objective effects. Thus, the reconstructed signal must be filtered.

Possibly, in this embodiment of the present invention, filtering can be performed with a low pass filter so that the pixel value in the boundary region changes more smoothly. For example, filtering is performed using a Gaussian filter. However, this is not limited in the present invention.

Possibly, in this embodiment of the present invention, the filtering may be performed using an Overlapped block motion compensation (OBMC) method for short. Motion-compensated prediction is performed on the filtered pixel using the motion vector of the affine MC block adjacent to the filtered pixel, and a weighted average is performed on the obtained motion-compensated prediction signal and the motion-compensated prediction signal obtained using the motion vector of the filtered pixel to obtain the final motion compensated prediction signal.

According to the video image decoding method in this embodiment of the present invention, the decoding device determines the size of the motion compensation affine sub-image sub-frame based on the determined difference of the motion vector and the motion vector precision of the affine image block, and the distance between the control points, and further corrects the size of the affine sub-image of the compensation image. motion based on the relationship between the size of the affine image block and the size of the affine sub-image sub-image of the motion compensation, so that the size of the affine sub-image of the motion compensation image is easy to realize and the complexity is reduced. It has been proved by experiments in the present invention that the complexity on the encoder side can be reduced by 50%, the complexity on the decoder side can be reduced by 70%, and the performance is maintained at 90% to 95%. Thus, the motion compensation prediction technology based on the affine model can be effectively applied in practice.

Possibly, the decoding method in the present invention may also be as follows:

parsing the bitstream to obtain the difference of the motion vector of the affine image block and the accuracy of the motion vector of the affine image block, where the pixels in the affine image block have the same affine transform parameter;

determining the size of an affine subblock of an image compensation of motion in an affine image block based on the motion vector difference, the motion vector precision, and the distance between control points in the affine image block, where the size includes the length in the horizontal direction and the length in the vertical direction, so that the length of the affine block the image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine sub-image of the motion compensation in the vertical direction; and control points are pixels used to determine the motion vector difference; and

performing decoding processing on the affine image block based on the size of the affine sub-block of the motion compensation image; or

parsing the bitstream to obtain affine image block information;

determining the size of the motion compensation image affine sub-block in the affine image block based on the affine image block information, where the size includes the length in the horizontal direction and the length in the vertical direction, the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-block motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-image compensation in the vertical direction; and

performing decoding processing on the affine image block based on the size of the affine subblock of the motion compensation image.

Thus, the decoder side does not need to recalculate, but directly use the parameter transmitted by the encoder side to the decoder side to determine the size of the affine sub-block of the motion compensation image. Thus, the calculation complexity is further reduced.

The above discussion describes in detail the video decoding method according to the embodiments of the present invention with reference to FIG. 1 - fig. 13. The above further illustrates that the way in which this solution is implemented on the decoder side and the encoder side is basically the same. For a clearer description, the following is a description of a video encoding method according to embodiments of the present invention with reference to FIG. 14 - fig. 16. It should be noted that since the related operations on the encoder side are essentially the same as those on the decoder side, in order to avoid repetition, the implementation details are not described again here.

14 is a flowchart of a video encoding method according to an embodiment of the present invention.

The method shown in FIG. 14 may be performed with an encoder. For example, similar to a decoder, an encoder typically includes a prediction block, a residual signal generation block, a transform block, a quantizer block, a dequantizer block, an inverse transform block, a reconstruction block, a filter block, a decoded image buffer, and an entropy coding block. . The entropy encoding block includes a conventional CABAC encoding/decoding block and a bypass encoding/decoding block. The prediction block includes an inter prediction block and an intra prediction block. The inter prediction block includes a motion estimator and a motion compensation block. In another embodiment, the encoder may include more or less or different functional components. Similar to the decoding method, the encoding method of the present invention is mainly applied to a method for performing inter prediction with an inter prediction block. For the detailed structure of each encoder function block, refer to the rules and description in the prior art, such as the H.265 standard. Detailed description in the present invention is omitted. In the present invention, only the following embodiment is used to describe the main process of adaptively determining the size of the motion compensation image affine sub-block and performing encoding in the method in this embodiment of the present invention.

In particular, as shown in FIG. 14, method 1400 includes the following steps:

S1410. Determine the difference of the motion vector of an affine image block.

S1430. Determine the accuracy of the motion vector of an affine image block.

S1450. Determine the size of an affine subblock of a motion compensation image in an affine image block based on the motion vector difference, the motion vector precision, the distance between control points in the affine image block, and the size of the affine image block, where the size includes the length in the horizontal direction and the length in the vertical direction , so that the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-image sub-block of the motion compensation in the vertical direction; and control points are pixels used to determine the motion vector difference.

S1470. Perform encoding processing on the affine image block based on the size of the affine subblock of the motion compensation image.

Specifically, the encoding device determines the motion vector difference of the affine image block based on the determined reference points, determines the motion vector accuracy of the affine image block, and determines the size of the affine sub-image sub-block of the motion compensation in the affine image block based on the determined motion vector difference and the motion vector accuracy, distance between breakpoints, and an affine image block size, where the size of the motion compensation affine image subblock is the size used to divide the affine image block, and the affine image block is divided into a plurality of affine motion compensation image subblocks with the same size based on the division size. The encoder then encodes the affine image block based on the affine subblock of the motion compensation image.

Thus, by executing the video encoding method in this embodiment of the present invention, the inter-picture prediction unit determines the size of the motion compensation affine image sub-block based on the determined motion vector difference and the motion vector precision of the affine image block, the distance between control points, and the size of the affine image block, split affine image block into a plurality of affine motion compensation image subblocks with one size based on the size of the affine motion compensation image subblock, and then performs encoding processing. Thus, in the encoding process, an affine sub-block of a motion compensation image with an appropriate size can be adaptively selected based on a change in the above parameters, encoding complexity is reduced, and encoding performance, that is, compression efficiency, is increased.

Preferably, S1410 includes:

determining a first motion vector difference component based on a difference between the motion vectors of the first reference point and the second reference point, which are located on the same horizontal line; and/or

determining a second motion vector difference component based on a difference between the motion vectors of the third reference point and the fourth reference point, which are located on the same vertical line.

There is a first horizontal distance between the first control point and the second control point, and there is a first vertical distance between the third control point and the fourth control point. It should be noted that "and/or" in the decision in step S1410 specifically refers to the following three different technical solutions:

Solution 1: A method for determining the motion vector difference of an affine image block based on three control points includes:

determining a first motion vector difference component based on a difference between the motion vectors of the first reference point and the second reference point, which are located on the same horizontal line, and determining a second motion vector difference component based on the difference between the motion vectors of the third reference point and the fourth reference point that are located on the same vertical line.

Solution 2: A method for determining a motion vector difference of an affine image block based on two control points in a horizontal direction includes: determining a first motion vector difference component based on a difference between the motion vectors of the first control point and the second control point, which are located on the same horizontal line.

Solution 3: A method for determining a motion vector difference of an affine image block based on two control points in the vertical direction includes: determining the second component of the motion vector difference based on the difference between the motion vectors of the third control point and the fourth control point, which are located on the same vertical line.

It should be understood that in this embodiment of the present invention, the first motion vector difference component is the horizontal motion vector difference component, and the second motion vector difference component is the vertical motion vector difference component. In Solution 1, the first motion vector difference component, the second motion vector difference component together constitute the motion vector difference of the affine image block. However, in Solution 2 and Solution 3, the first motion vector difference component, the second motion vector difference component independently reflect the motion vector difference of the affine image block. In other words, in Solution 1, the motion vector difference of the affine image block includes a first motion vector difference component, a second motion vector difference component; in solution 2, the motion vector difference of the affine image block includes the first motion vector difference component; and in solution 3, the motion vector difference of the affine image block includes a third component of the motion vector difference. It should also be understood that in this embodiment of the present invention, the first, second, third and fourth are only used to distinguish pixels and should not constitute any limitation on the protection scope of the present invention. For example, the first reference point may also be referred to as the second reference point, and the second reference point may be referred to as the first reference point.

Perhaps, in this embodiment of the present invention, the first control point and the third control point are the same pixels. More specifically, three control points are selected, and one of the three control points is on the same horizontal line as one of the other two control points, and is located on the same vertical line as the last control point.

Perhaps, in this embodiment of the present invention, the first control point, the second control point, the third control point and the fourth control point are the vertices of the affine block. The four vertices in an affine block are assumed to be respectively labeled top left, bottom left, top right, and bottom right.

In solution 1, any three vertices in two orthogonal directions can be chosen to determine the difference in the motion vector of the affine block. For example, the top left vertex, bottom left vertex, and top right vertex may be selected, or the top left vertex, top right vertex, and bottom right vertex may be selected.

In solution 2, any two vertices located on the same horizontal line can be selected to determine the difference in the motion vector of the affine block. For example, the top left vertex and top right vertex may be selected, or the bottom left vertex and bottom right vertex may be selected.

In solution 3, any two vertices located on the same vertical line can be selected to determine the difference in the motion vector of the affine block. For example, the top left vertex and bottom left vertex may be selected, or the top right vertex and bottom right vertex may be selected. This is not limited in the present invention.

Alternatively, in this embodiment of the present invention, the difference between the components that are the motion vectors of the first reference point and the second reference point, and which are in one of two directions (horizontal direction and vertical direction, or referred to as the x direction and direction y) as the first difference component of the motion vector of the affine block, and the difference between the components that have the motion vectors of the third reference point and the fourth reference point in either of the two directions as the second difference component of the motion vector of the affine block can be determined. The method for determining the motion vector difference is applied to solutions 1, 2, and 3. Alternatively, according to the actual needs of encoding complexity and encoding performance, the values between two differences between the motion vector components of the first reference point and the second reference point in two directions may be is selected and determined as the first component, and a value between two differences between the motion vector components of the third reference point and the fourth reference point in two directions may be selected and determined as the second component. This is not limited in the present invention.

The motion vector of each from the first waypoint to the fourth waypoint typically includes two components, ie, a horizontal component and a vertical component. Thus, in the preceding steps, determining the first motion vector difference component based on the difference between the motion vectors of the first reference point and the second reference point, which are located on the same horizontal line, and determining the second motion vector difference component based on the difference between the motion vectors of the third reference point and the fourth control point, which are located on the same vertical line, can be specifically implemented as follows:

determining a first horizontal difference component and a first vertical difference component between the motion vectors of the first reference point and the second reference point, and determining more than one of the first horizontal difference component and the first vertical difference component as the first component of the motion vector difference; and

respectively, determining a second horizontal difference component and a second vertical difference component between the motion vectors of the third reference point and the fourth reference point, and determining more than one of the second horizontal difference component and the second vertical difference component as the second motion vector difference component.

Perhaps, in this embodiment of the present invention, the motion vector difference is obtained based on an affine model, that is, the motion vector differences are obtained by determining the affine transformation parameters of a pixel in an affine image block. It should be noted that pixels in an affine image block have the same affine transformation parameter.

Accordingly, the above step of determining the first horizontal difference component and the first vertical difference component between the motion vectors of the first reference point and the second reference point is, namely: determining the first horizontal difference component and the first vertical difference component based on the affine transformation parameter and the first horizontal distance; and the preceding step of determining the second horizontal difference component and the second vertical difference component between the motion vectors of the third reference point and the fourth reference point is precisely: determining the second horizontal difference component and the second vertical difference component based on the affine transform parameter and the first vertical distance.

It should be understood that the sequence numbers of the above processes do not indicate the sequence of execution. The sequence of execution of the processes should be determined based on the functions and internal logic of the processes, and should not represent any limitation on the processes of implementing this embodiment of the present invention.

Possibly, in this embodiment of the present invention, the affine transform model may be a known 6-parameter affine transform model, or may be a prior art 4-parameter or 2-parameter affine transform model. The 6-parameter affine transformation model is mainly applied to the scenario in solution 1 based on three control points, and the 4-parameter affine transformation model is mainly applied to the scenarios in solution 2 and solution 3 based on two control points. For specific content, see the descriptions of the decoding method in this specification.

Perhaps, in this embodiment of the present invention, the motion vector of the first reference point, the motion vector of the second reference point, the motion vector of the third reference point, and the motion vector of the fourth reference point can be determined to determine the motion vector difference of the affine image block.

Accordingly, in the scenario in Solution 1, the difference between the horizontal motion vector component of the first reference point and the horizontal motion vector component of the second reference point is determined as the first component of the horizontal difference; the difference between the vertical component of the motion vector of the first reference point and the vertical component of the motion vector of the second reference point is determined as the first component of the vertical difference; the difference between the horizontal motion vector component of the third reference point and the horizontal motion vector component of the fourth reference point is determined as the second component of the horizontal difference; and the difference between the vertical motion vector component of the third reference point and the vertical motion vector component of the fourth reference point are determined as the second component of the vertical difference.

In the scenario in Solution 2, the difference between the horizontal motion vector component of the first control point and the horizontal motion vector component of the second control point is defined as the first component of the horizontal difference; and the difference between the vertical motion vector component of the first reference point and the vertical motion vector component of the second reference point is defined as the first vertical difference component.

In the scenario in Solution 3, the difference between the horizontal motion vector component of the third control point and the horizontal motion vector component of the fourth control point is defined as the second component of the horizontal difference; and the difference between the vertical motion vector component of the third reference point and the vertical motion vector component of the fourth reference point are determined as the second component of the vertical difference.

In particular, the motion vectors of the control points can be directly determined, and the first component and the second component of the motion vector difference of the affine block can be directly obtained by calculating the difference between the components of the motion vectors.

Perhaps, in this embodiment of the present invention, the first control point and the second control point are two adjacent pixels, and the third control point and the fourth control point are two adjacent pixels. In other words, the control points are not necessarily chosen from the four corner points, also referred to as vertices, in the affine image block, but can be points that are at any position in the affine image block and are jointly set on the encoder side and the decoder side, as explicitly stated. in a bitstream using signaling, or derived from checkpoint-based estimation or duplication in an affine picture block environment. In this case, the definition of the first horizontal difference component, the first vertical difference component, the second horizontal difference component, and the second vertical difference component is exactly as follows:

A motion vector of the first pixel, a motion vector of the second pixel, and a motion vector of the third pixel are determined, where the first pixel, the second pixel, and the third pixel are pixels that do not overlap with each other; determining a second horizontal distance and a second vertical distance between the first pixel and the second pixel; determining a third horizontal distance and a third vertical distance between the first pixel and the third pixel; and determining a first horizontal difference component, a first vertical difference component, a second horizontal difference component, and a second vertical difference component based on the motion vector of the first pixel, the motion vector of the second pixel, the motion vector of the third pixel, the second horizontal distance, the second vertical distance, the third horizontal distance, and third vertical distance. Solution 1 describes an example based on three breakpoints. As evident from the above example, the difference between solution 2 or solution 3 based on two control points and solution 1 based on three control points is: use only the first component of the difference or the second component of the difference in solution 1 as the difference of the motion vector of the affine image based on two control points.

Possibly, in this embodiment of the present invention, the encoder may obtain the motion vectors of all control points through motion estimation and searching, may obtain the motion vectors of all control points based on the surrounding image block, or may calculate the motion vectors of all control points based on an affine transformation model. Alternatively, the encoder may obtain the motion vectors of some control points through motion estimation and searching, and obtain the motion vectors of other control points based on the neighboring image block. Alternatively, the encoder may obtain the motion vectors of some control points using affine motion estimation and search, and calculate the motion vectors of other control points based on the affine transformation model. Alternatively, the encoder may derive the motion vectors of some control points based on an adjacent image block and calculate the motion vectors of other control points based on an affine transformation model. However, this is not limited in the present invention.

Perhaps, in S1430, the first set value can be determined as the motion vector precision of the affine image block; or, the motion vector accuracy of the affine image block may be determined based on a feature of the adjacent image block of the affine image block. An adjacent image block is an image block next to an affine image block in space and/or time.

Perhaps, as shown in Fig. 15, specifically, S1450 includes the following steps.

S1501. To proportionally adjust the first horizontal distance based on the ratio of the motion vector accuracy to the first motion vector difference component to obtain the length of the affine motion compensation image subblock in the horizontal direction; and determine the length of the affine image block in the horizontal direction, whether it is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction direction, adjust the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal motion compensation affine image subblock.

In particular, the ratio of the motion vector accuracy to the first motion vector difference component is the ratio of the motion vector accuracy to the first motion vector difference component. The product of the ratio and the first horizontal distance is the length of the affine sub-image compensation motion in the horizontal direction, and can also be referred to the initial length in the horizontal direction. It may be understood that in some cases, the initial length of the horizontal motion compensation affine sub-frame of the image may not be an integer. In this case, the initial length must be rounded off. To ensure accuracy, rounding down is usually performed. More specifically, the decimal part is directly discarded. For example, the initial length value is 8.35 (that is, a length of 8.35 unit pixels), and 8 is directly used as the length of the affine subblock of the horizontal motion compensation image. Of course, the principle of rounding up can alternatively be used. More specifically, if the fractional part is less than 5, rounding down is performed; and, if the fractional part is greater than 5, rounding up is performed. For example, when the length is calculated as 8.86, 9 is used as the length. To ensure performance, the minimum value of the initial length should not be less than the specified value, for example, not less than the specified value, such as 2, 4, 8 or 2 ^N . The specified value is usually set to 4. The initial length should also not be greater than the length of the affine image block in the horizontal direction. Whether the initial length in the horizontal direction is correct or the rounded initial length in the horizontal direction depends on whether the length of the affine image block in the horizontal direction is divisible by the initial length. If the length of the affine image block in the horizontal direction is divided by the initial length, then this indicates whether the affine image block can be divided into an integer number of sub-blocks in the horizontal direction based on the initial length in the horizontal direction or the rounded initial length in the horizontal direction. The correction method is generally a method in which the initial length in the horizontal direction or the rounded initial length in the horizontal direction is gradually increased/decreased. More specifically, if it is determined that the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, then the length of the affine sub-image of the compensation image in the horizontal direction is corrected, that is, increase/decrease by one or several unit lengths, and then cycle through the previous decision until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal motion compensation. The length unit is typically one block of pixels. More specifically, the length of the affine subblock of the motion compensation image in the horizontal direction is corrected, that is, increased/decreased by 1. Without loss of generality, in other words, the length of the affine subblock of the motion compensation image in the horizontal direction is corrected, that is, increased/decreased by a fixed step size. The step size is a correction value, that is, a base block or amplitude used to correct the length of the affine sub-block of the horizontal direction motion compensation image in each cycle. The step size is typically the aforementioned one block of pixels, or may be another value such as two blocks of pixels. It is possible that the size value/adjustment step used in the above adjustment may change depending on the ratio of the original length of the horizontal motion compensation affine sub-image sub-block. For example, when the initial length of the affine subblock of the horizontal motion compensation image is an odd number, an odd step size is selected. For example, an odd number of unit lengths, such as one unit length, is used as the step size for adjustment. When the initial length of the affine subblock of the horizontal motion compensation image is an even number, an even step size is selected. For example, an even number of unit lengths is used as the step size.

In addition, whether the adjustment is an increase or decrease may be selected according to an agreement between the decoder side and the encoder side, may be explicitly indicated on the encoder side in the bitstream, and is selected on the decoder side based on explicit indication, may be inferred using accuracy of the motion vector of the surrounding neighboring block, or may be determined based on the relation of the ratio between the length of the affine image block in the horizontal direction and the length of the affine image sub-block of the motion compensation image in the horizontal direction. For example, if the ratio is less than a threshold value, decrease is selected; otherwise, increase is selected.

To simplify the algorithm and ensure the efficiency of encoding and decoding, if it is determined that the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, the length of the affine sub-image sub-image of the compensation motion in the horizontal direction is corrected, that is increase by one predetermined length unit, and then the preceding decision will be cycled until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the motion compensation in the horizontal direction.

Further, in order to increase efficiency and reduce operation time, the length of the affine sub-image block of the horizontal direction motion compensation may be brought to the nearest value by which the length of the affine image block in the horizontal direction is divided. For example, the length of the affine image subblock of the horizontal direction motion compensation image is 7, while the length of the affine image block in the horizontal direction is 16. In this case, the length of the affine image subblock of the horizontal direction motion compensation image is increased by 1 and changed to 8, with by which 16 is divided. Similarly, if the length of the horizontal motion compensation affine image subblock is 5, the length of the horizontal motion compensation affine image subblock is reduced by 1 to change to 4, by which 16 is divided. If the difference between the length of the horizontal direction motion compensation affine image sub-block and two adjacent values by which the horizontal direction affine image length is divided is equal, for example, the length of the horizontal direction motion compensation affine image sub-block is 6, the length of the horizontal direction affine image block is direction is 16, and two values that are adjacent with a length of 6 and by which 16 is divided are 4 and 8, respectively, the correction can be selected according to the requirements of the encoder side and the decoder side. For example, if low latency is desired, then gain correction is performed. If high quality is required, a reduction correction is performed. This is not limited in the present invention.

, Wmax - это возможно максимальная длина аффинного блока изображения в горизонтальном направлении. Определяют интервал, в котором находится длина аффинного субблока изображения компенсации движения в горизонтальном направлении, и значение нижнего предела или значение верхнего предела интервала, в котором находится длина аффинного субблока изображения компенсации движения в горизонтальном направлении, используются как окончательно определенная длина аффинного субблока изображения компенсации движения в горизонтальном направлении. Правило получения значения нижнего предела или значения верхнего предела, относится к соглашению между стороной кодера и стороной декодера. Верхнее предельное значение или нижнее предельное значение приближенного интервала может быть явно указано, используя поток битов, или может быть выбрано посредством округления. С помощью приведенных выше таблиц, окончательная длина аффинного субблока изображения компенсации движения в горизонтальном направлении может быть определена с помощью таблицы поиска на основе начальной длины аффинного субблока изображения компенсации движения в горизонтальном направлении. Таким образом, этап корректировки длины аффинного субблока изображения компенсации движения в горизонтальном направлении на этапе S1501 можно заменить, сложность системы снижается, и время вычисления уменьшается.Further, in step S1501, when it is determined that the length of the affine picture block in the horizontal direction is not an integer multiple of the length of the affine picture sub-picture compensation in the horizontal direction, in order to further simplify the algorithm, the same table can be separately set at the encoder side, and side of the decoder, and the possible maximum length of the affine image block in the horizontal direction is divided into a plurality of intervals, using the allowable minimum divisor as the initial value, for example, 2 ^N . Interval range 2 ^N+i , where

, Wmax is the possible maximum length of an affine image block in the horizontal direction. The interval in which the length of the affine sub-frame of the motion compensation image in the horizontal direction is located is determined, and the value of the lower limit or the value of the upper limit of the interval in which the length of the affine sub-frame of the motion compensation image in the horizontal direction lies is used as the finally determined length of the affine sub-frame of the motion compensation image in horizontal direction. The rule for obtaining a lower limit value or an upper limit value refers to an agreement between the encoder side and the decoder side. The upper limit value or the lower limit value of the approximate interval may be explicitly indicated using the bitstream, or may be selected by rounding. With the above tables, the final length of the horizontal direction motion compensation affine image subblock can be determined with a lookup table based on the initial length of the horizontal direction motion compensation affine image subblock. Thus, the step of adjusting the length of the affine sub-frame of the horizontal direction motion compensation image in step S1501 can be replaced, the complexity of the system is reduced, and the calculation time is reduced.

It is possible to further reduce the computational complexity in the decision in step S1501, and not only if the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image sub-block of the horizontal motion compensation, it is also determined whether the length of the affine image sub-image block is Compensation for movement in the horizontal direction is an integer multiple of the specified length S, which facilitates the calculation. More specifically, it must be determined whether the length of the affine subblock of the horizontal motion compensation image meets the following conditions:

(1) The length of the affine image block in the horizontal direction is an integer multiple of the length of the affine sub-image block of the horizontal motion compensation; and (2) the length of the affine subblock of the horizontal motion compensation image is an integer multiple of the predetermined length S.

If the length of the affine motion compensation image subblock in the horizontal direction does not meet the two conditions, then the length of the affine motion compensation image subblock in the horizontal direction is adjusted so that the length of the affine subblock of the motion compensation image in the horizontal direction satisfies condition (1) and condition (2). More specifically, the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine image sub-image compensation in the horizontal direction, and the corrected length of the affine image sub-image of the compensation in the horizontal direction is an integer multiple of the predetermined length. The given length S is typically 4, or may be a value such as 8, 16, or 32. In other words, S = 2 ⁿ , where n is zero or a positive integer. This solution is applicable to an affine image block in any form. When an affine image block is divided according to the rule of a Quadtree (Qual Tree), provided that any of the above defining conditions (1) and (2) is satisfied, another condition may be satisfied. Thus, only one of the defining conditions must be activated/applied.

The length amplitude limiting operation may need to be performed, which is the length of the affine subblock of the horizontal motion compensation image, as determined by the above. More specifically, a comparison method is used to allow M to be within the range from the lower limit Tm to the upper limit W. Tm is a predetermined value and is generally equal to 4, and W is the length of the affine image block in the horizontal direction.

S1503. To proportionally adjust the first vertical distance based on the ratio of the motion vector accuracy to the second motion vector difference component to obtain the length of the affine motion compensation image subblock in the vertical direction; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-image block in the vertical direction and, if the length of the affine image block in the vertical direction is not an integer, a multiple of the length of the affine image sub-image sub-block of the vertical motion compensation , adjust the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the vertical direction motion compensation affine image subblock.

In particular, the ratio of the motion vector accuracy to the second motion vector difference component is the ratio of the motion vector accuracy to the second motion vector difference component. The product of the ratio and the first vertical distance is the length of the affine subblock of the vertical motion compensation image, and can also be referred to the initial length in the vertical direction. Obviously, the initial length of the affine subblock of the vertical motion compensation image may not be an integer, in some cases. In this case, the initial length must be rounded off. To ensure accuracy, rounding down is usually performed. More specifically, the decimal part is directly discarded. For example, the initial length value is 8.35 (that is, a length of 8.35 unit pixels), and 8 is directly used as the length of the vertical motion compensation image affine subblock. Of course, the principle of rounding may alternatively be used. More specifically, if the fractional part is less than 5, then rounding down is performed; and, if the fractional part is greater than 5, then rounding up is performed. For example, when the length is calculated as 8.86, then 9 is used as the length. To ensure performance, the minimum value of the initial length should not be less than the specified value, for example, not less than the specified value, such as 2, 4, 8 or 2 ^N . The specified value is usually set to 4. The initial length should also not be greater than the length of the affine image block in the vertical direction. Whether the initial length in the vertical direction is correct or the rounded initial length in the vertical direction depends on whether the length of the affine image block in the vertical direction is divisible by the initial length. If the affine image block length in the vertical direction is divided by the initial length, then this indicates whether the affine image block can be divided into an integer number of sub-blocks in the vertical direction based on the initial length in the vertical direction or the rounded initial length in the vertical direction. The correction method is generally a method in which the initial length in the vertical direction or the rounded initial length in the vertical direction is gradually increased/decreased. More specifically, if it is determined that the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine sub-image sub-block of the vertical motion compensation image, then the length of the affine sub-image sub-block of the vertical motion compensation image is corrected, that is, increased/decreased, by one or more unit lengths, and then cycle through the previous decision until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. The length unit is typically one block of pixels. More specifically, the length of the vertical direction motion compensation affine sub-image subblock is corrected, that is, increased/decreased by 1. Without loss of generality, in other words, the length of the vertical motion compensation affine sub-image sub-block is adjusted, that is, increased/decreased by a fixed step size. The step size is an adjustment value, that is, a basic block or amplitude, used to correct the length of the affine sub-block of the vertical motion compensation image in each cycle. The step size is typically referred to as one block of pixels, or may be another value, such as two blocks of pixels. It is possible that the size value/correction step used in the above correction may vary depending on the ratio of the initial length of the vertical motion compensation affine sub-image sub-block. For example, when the initial length of the affine subblock of the vertical motion compensation image is an odd number, an odd step size is selected. For example, an odd number of length units, such as one length unit, is used as the step size for adjustment. When the initial length of the affine subblock of the vertical motion compensation image is an even number, an even step size is selected. For example, an even number of unit lengths is used as the step size.

In addition, whether the adjustment is an increase or decrease can be selected according to an agreement between the decoder side and the encoder side, can be explicitly indicated on the encoder side in the bitstream, and is selected on the decoder side based on explicit indication, can be obtained using precision the motion vector of the surrounding neighboring block, or may be determined based on the relation of the ratio between the length of the affine image block in the vertical direction and the lengths of the affine sub-image subblock of the motion compensation in the vertical direction. For example, if the ratio is less than a threshold value, decrease is selected; otherwise, increase is selected.

To simplify the algorithm and ensure the efficiency of encoding and decoding, if it is determined that the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine sub-image sub-block of the vertical motion compensation image, then the length of the affine sub-image sub-block of the image compensation in the vertical direction is corrected, then is incremented by one predetermined length unit, and then the said decision is cycled until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation.

Further, in order to improve efficiency and reduce operation time, the length of the affine sub-image block of the vertical direction motion compensation can be brought to the nearest value by which the length of the affine image block in the vertical direction is divided. For example, the length of the vertical direction motion compensation affine image subblock is 7, while the length of the vertical direction affine image block is 16. In this case, the length of the vertical direction compensation affine image subblock is increased by 1 to change to 8, with which 16 is divisible. Similarly, if the length of the vertical motion compensation affine sub-image sub-frame is 5, then the length of the vertical motion compensation affine sub-frame is reduced by 1 to change to 4, by which 16 is divided. If the difference between the length of the vertical direction motion compensation affine image block and two adjacent values by which the length of the vertical direction affine image block is divided is equal, for example, the length of the vertical direction motion compensation affine image block is 6, the length of the affine image block in vertical direction is 16, and two values that are adjacent with a length of 6 and by which 16 is divided equal to 4 and 8, respectively, the correction can be selected according to the requirements of the encoder side and the decoder side. For example, if low latency is desired, then gain correction is performed. If high quality is required, then select the reduction adjustment. This is not limited in the present invention.

и Wmax это возможно максимальная длина аффинного блока изображения в вертикальном направлении. Определяют интервал, в рамках которого находится длина аффинного субблока изображения компенсации движения в вертикальном направлении, и значение нижнего предела или значение верхнего предела интервала, в рамках которого находится длина аффинного субблока изображения компенсации движения в вертикальном направлении, используются как окончательно определенная длина аффинного субблока изображения компенсации движения в вертикальном направлении. Правило для получения значения нижнего предела или значения верхнего предела установлено по соглашению между стороной кодера и стороной декодера. Верхнее предельное значение или нижнее предельное значение приближенного интервала может быть явно указано, используя поток битов, или может быть выбрано согласно округления. С помощью приведенных выше таблиц, окончательная длина аффинного субблока изображения компенсации движения в вертикальном направлении может быть определена с помощью таблицы поиска на основе начальной длины аффинного субблока изображения компенсации движения в вертикальном направлении. Таким образом, этап корректировки длины аффинного субблока изображения компенсации движения в вертикальном направлении на этапе S1503 можно заменить, сложность системы снижается, и время вычисления уменьшается.Further, in step S1503, when it is determined that the affine image block length in the vertical direction is not an integer multiple of the length of the vertical motion compensation affine sub-image block, in order to further simplify the algorithm, the same table can be separately set at the encoder side, and side of the decoder, and the possible maximum length of the affine image block in the vertical direction is divided into a plurality of intervals, using the allowable minimum divisor as the initial value, for example, 2 ^N . The interval is 2 ^N+i , where

and Wmax is the possible maximum length of an affine image block in the vertical direction. The interval within which the length of the affine sub-image compensation in the vertical direction lies, and the value of the lower limit or the value of the upper limit of the interval within which the length of the affine sub-image of the compensation in the vertical direction lies are used as the finally determined length of the affine sub-image of the compensation image. movement in the vertical direction. The rule for obtaining the lower limit value or the upper limit value is set by agreement between the encoder side and the decoder side. The upper limit value or the lower limit value of the approximate interval may be explicitly specified using the bitstream, or may be selected according to rounding. With the above tables, the final length of the affine sub-image sub-frame of the vertical motion compensation image can be determined with a lookup table based on the initial length of the affine sub-frame of the vertical motion compensation image. Thus, the step of correcting the length of the vertical motion compensation affine subblock image in step S1503 can be replaced, the system complexity is reduced, and the calculation time is reduced.

Alternatively, in order to further reduce the computational complexity in the decision, at step S1503, not only is it determined if the length of the affine image block in the vertical direction is an integer multiple of the length of the affine sub-image sub-block of the vertical motion compensation image, but it is determined whether the length of the affine image block in the vertical direction is image compensation motion in the vertical direction is an integer multiple of the specified length S, which will reduce the complexity of calculations. More specifically, it is determined whether the length of the affine sub-block of the vertical motion compensation image meets the following conditions:

(1) The length of the affine image block in the vertical direction is an integer multiple of the length of the affine sub-image block of the vertical motion compensation; and (2) the length of the affine subblock of the vertical motion compensation image is an integer multiple of the given length S.

If the length of the vertical direction motion compensation image affine subblock does not meet two conditions, then the length of the vertical direction motion compensation image affine subblock is adjusted so that the length of the vertical direction motion compensation image affine subblock satisfies condition (1) and condition (2). More specifically, the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the vertical motion compensation affine image subblock, and the corrected length of the affine image subblock of the vertical motion compensation image is an integer multiple of the predetermined length. The given length S is typically 4, or may be a value such as 8, 16, or 32. In other words, S = 2 ⁿ , where n is zero or a positive integer. This solution is applicable to an affine image block in any form. When an affine image block is partitioned according to a Quadtree rule, provided that any of the above defining conditions (1) and (2) is satisfied, another condition may be satisfied. Thus, only one of the defining conditions must be activated/applied.

Alternatively, an operation of limiting the amplitude along the length, which is the length of the affine subblock of the vertical motion compensation image, as defined above, should be performed. More specifically, a comparison method is used to ensure that M is within the range from the lower limit Tm to the upper limit W. Tm is a predetermined value and is generally 4, and W is the length of the affine image block in the vertical direction.

Description of specific implementations of determining and adjusting the size of the affine sub-block of the motion compensation image can be understood with reference to the specific description of the decoding method in this document.

That is, according to the video encoding method in this embodiment of the present invention, the encoding apparatus determines the size of the motion compensation affine image sub-block based on the determined motion vector difference and the motion vector precision of the affine image block, the distance between control points, and further corrects the size of the affine image sub-block motion compensation based on the relationship between the size of the affine image block and the size of the affine sub-image of the motion compensation image, so that the size of the affine sub-image of the motion compensation image is easy to realize and the complexity is reduced. It has been proved by experiments in the present invention that the complexity on the encoder side can be reduced by 50%, the complexity on the decoder side can be reduced by 70%, and the performance is maintained at 90% to 95%. Thus, the motion compensation prediction technology based on the affine model can be effectively applied in practice.

Alternatively, as shown in FIG. 16, S1470 includes the following steps:

S1601. Determine the motion vector of each affine subblock of the motion compensation image in the affine subblocks of the motion compensation image.

S1603. Determine the motion compensation prediction signal of each affine motion compensation image subblock based on the motion vector of each affine motion compensation image subblock.

S1605. Encode each affine sub-block of the motion compensation image based on the motion-compensated prediction signal of each affine sub-block of the motion compensation image.

It should be understood that in this embodiment of the present invention, each affine MC block may include a plurality of pixels, and a motion vector of pixels from each affine MC block must be selected as the motion vector of the affine MC block.

Alternatively, in S1601, the motion vector of the pixel at the center position of the affine MC block may be selected as the motion vector of the affine MC block, the average of the motion vectors of all pixels in the affine MC block may be used as the motion vector of the affine MC block, or may be the motion vector of the top left pixel in the affine MC block is selected as the motion vector of the affine MC block. However, this is not limited in the present invention.

Alternatively, in this embodiment of the present invention, the boundary pixel signal in each affine sub-frame of a motion compensation image is filtered, and the boundary pixel is one or more rows of pixels at the boundary of each affine sub-frame of a motion compensation image.

Alternatively, in this embodiment of the present invention, the boundary pixel signal includes a motion-compensated prediction signal and/or a reconstructed signal, and the reconstructed signal is the sum of the motion-compensated prediction signal and the reconstructed residual signal.

The recovered signal is typically obtained by summing the motion-compensated prediction signal and the recovered residual signal. Lossy coding is usually performed on the residual signal. This results in a distortion of the reconstructed residual signal with respect to the original residual signal. Pixel distortion at the boundaries of adjacent affine MC blocks can be different. For example, a pixel on the right side in the first affine MC block becomes larger due to distortion, and a pixel on the left side in an affine MC block that is adjacent to and to the right of the first affine MC block becomes smaller due to distortion. This causes a break in the boundary values of the reconstructed pixels in the affine MS blocks, and also distorts the subjective and objective effects. Thus, the reconstructed signal must be filtered.

Alternatively, in this embodiment of the present invention, filtering can be performed with a low pass filter so that the pixel value in the boundary region changes more smoothly. For example, filtering is performed using a Gaussian filter. However, this is not limited in the present invention.

Alternatively, in this embodiment of the present invention, the filtering may be performed using an Overlapped block motion compensation (OBMC) method for short. Motion-compensated prediction is performed to filter the pixel using the motion vector of the affine MC block adjacent to the filtered pixel, and weighted averaging is performed on the obtained motion-compensated prediction signal and the motion-compensated prediction signal obtained using the motion vector of the filtered pixel to obtain the final motion compensated prediction signal.

According to the video coding method in this embodiment of the present invention, the decoding apparatus determines the size of the compensation motion affine sub-image sub-frame based on the determined motion vector difference and the motion vector precision of the affine picture block, and the distance between control points, and further corrects the size of the compensation affine sub-frame image motion based on the relationship between the size of the affine image block and the size of the affine sub-image sub-image of the motion compensation, so that the size of the affine sub-image of the motion compensation image is easy to realize and the complexity is reduced. It has been proved by experiments in the present invention that the complexity on the encoder side can be reduced by 50%, the complexity on the decoder side can be reduced by 70%, and the performance is maintained at 90% to 95%. Thus, the motion compensation prediction technology based on the affine model can be effectively applied in practice.

Decoding equipment 1700 for implementing the present invention is described below with reference to FIG. 17. Equipment 1700 for decoding includes:

a motion vector difference determination module 1710, configured to determine a motion vector difference of the affine image block;

a motion vector accuracy determination unit 1730, configured to determine the motion vector accuracy of the affine image block;

an image sub-block size determining module 1750, configured to determine the size of an affine sub-image sub-block of a motion compensation in the affine image block based on a motion vector difference, a motion vector accuracy, a distance between control points in the affine image block, and an affine image block size, where the size includes a length in the horizontal direction and a length in the vertical direction, such that the length of the affine image block in the horizontal direction is an integer multiple of the length of the motion compensation affine image sub-image in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length an affine subblock of a vertical motion compensation image; and control points are pixels used to determine the motion vector difference; and

a decoding unit 1770, configured to perform decoding processing on the affine image block based on the size of the affine subblock of the motion compensation image.

Specifically, the decoding equipment determines the motion vector difference of the affine image block based on the determined reference points, determines the motion vector accuracy of the affine image block, and determines the size of the affine image subblock of the motion compensation image in the affine image block based on the determined motion vector difference and the motion vector accuracy, distance between the breakpoints and the affine image block size, where the size of the motion compensation affine image subblock is the size used to divide the affine image block, and the affine image block is divided into a plurality of affine motion compensation image subblocks with the same size based on the partition size. Then, the decoding equipment decodes the affine image block based on the affine subblocks of the motion compensation image. The decoding equipment modules implement the functions of various steps in the decoding method of the present invention. More specifically, the modules implement the functions of all steps of the decoding method in the method of the present invention, for example, steps S710 to S740 before extensions and variations of these steps. For more information, reference is made to the description of the decoding method in this specification. For brevity, the details are not described here again.

The following describes in detail the encoding equipment according to an embodiment of the present invention with reference to FIG. 18. Encoding equipment 1800 includes:

a motion vector difference determination module 1810, configured to determine a motion vector difference of the affine image block;

a motion vector accuracy determination module 1830, configured to determine the motion vector accuracy of the affine image block;

an image sub-block size determination module 1850, configured to determine the size of an affine image sub-block of a motion compensation image in the affine image block based on a motion vector difference, motion vector precision, distance between control points in the affine image block, and a size of the affine image block, where the size includes a length in the horizontal direction and a length in the vertical direction, such that the length of the affine image block in the horizontal direction is an integer multiple of the length of the motion compensation affine image sub-image in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length an affine subblock of a vertical motion compensation image; and control points are pixels used to determine the motion vector difference; and

an encoding unit 1870, configured to perform encoding processing on the affine image block based on the size of the affine sub-block of the motion compensation image.

Specifically, the encoding equipment determines the motion vector difference of the affine image block based on the determined reference points, determines the motion vector accuracy of the affine image block, and determines the size of the affine image sub-block of the motion compensation image in the affine image block based on the determined motion vector difference and the motion vector accuracy, the distance between the control points and the size of the affine image block, where the size of the affine motion compensation image subblock is the size used to divide the affine image block, and the affine image block is divided into a plurality of affine motion compensation image subblocks with the same size based on the partition size. The encoding equipment then encodes the affine image block based on the affine sub-blocks of the motion compensation image. Encoding equipment modules implement the functions of the various steps of the encoding method according to the present invention. More specifically, the modules implement the functions of all steps of the encoding method in the method of the present invention, such as steps S1410 to S1470 and extensions and variations of these steps. For more information, reference is made to the description of the encoding method in this document. For brevity, the details are not described in this description again.

As shown in FIG. 19, an embodiment of the present invention further provides a decoding apparatus 1900 including a processor 1910, a memory 1930, and a system bus 1950. The processor and memory are connected via a system bus. The memory is configured to store instructions. The processor is configured to execute instructions stored in memory. The memory stores the program code. The processor may call the program code stored in memory to perform all the steps and extend the steps in the decoding method 700 according to the present invention, that is, all the methods described in FIG. 7 - fig. thirteen.

For example, the processor may call the program code stored in memory to perform full decoding operations in the method of the present invention:

determining a motion vector difference of the affine image block;

determining the accuracy of the motion vector of the affine image block;

determining the size of an affine subblock of a motion compensation image in an affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, where the size includes a length in the horizontal direction and a length in the vertical direction, so that the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image subblock of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation in the vertical direction; and control points are pixels used to determine the motion vector difference; and

performing decoding processing on the affine image block based on the size of the affine subblock of the motion compensation image.

In the decoding process, an image sub-block with an appropriate size can be selected, and the image size of the sub-block is adjusted so that the affine image block is divided into a plurality of affine motion compensation image sub-blocks with the same size. Thus, the decoding complexity is reduced and the decoding performance is improved.

It should be understood that in this embodiment of the present invention, the processor 1910 may be a central processing unit (Central Processing Unit, "CPU" for short). Alternatively, processor 1910 may be another general purpose processor, digital signal processor (DSP), application specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete logic element or transistorized logic device, discrete hardware component or the like. A general purpose processor may be a microprocessor, or the processor may be any conventional processor or the like.

Memory 1930 may include read-only memory and random access memory, and provide instructions and data to processor 1910. Part of memory 1930 may further include non-volatile random access memory. For example, memory 1930 may additionally store device type information.

In addition to the data bus, system bus 1950 may further include a power bus, a control bus, a status signal bus, and the like. However, for a clear description, the various types of tires in the drawing are marked as 1950 system bus.

During implementation, the steps of the methods described above may be implemented by integrated logic circuitry in processor 1910 or by software instructions. The steps of the methods described with reference to embodiments of the present invention may be directly executed in a hardware processor, or may be performed using a combination of processor hardware and a software module. The program module may be located on a storage medium known in the art, such as random access memory, flash memory, read-only memory, read-only programmable memory, electrically erasable programmable memory, or a register. The storage medium resides in memory 1930 and the processor 1910 reads information from memory 1930 and performs the steps of the above methods in conjunction with the processor hardware.

As shown in FIG. 20, an embodiment of the present invention further provides an encoder 2000 including a processor 2010, a memory 2030, and a system bus 2050. The processor 2010 and memory 2030 are coupled via a system bus 2050. The memory 2030 is configured to store instructions. The processor 2010 is configured to execute an instruction stored in the memory 2030. The memory 2030 of the encoder 2000 stores program code. The processor 2010 may call the program code stored in the memory 2030 to perform all of the steps and extensions of the steps in the encoding method 1400 according to the present invention, that is, all of the method steps described in FIG. 14 - fig. 16. For example, processor 2010 may call program code stored in memory 2030 to perform the following operations:

determining the difference of the motion vector of the affine image block;

determining the accuracy of the motion vector of the affine image block;

determining the size of an affine subblock of a motion compensation image in an affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, where the size includes a length in the horizontal direction and a length in the vertical direction, so that the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image subblock of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation in the vertical direction; and control points are pixels used to determine the motion vector difference; and

performing encoding processing on the affine image block based on the size of the affine image sub-block of the motion compensation image.

In the encoding process, an image sub-block with an appropriate size can be selected and the size of the image sub-block is adjusted so that the affine image block is divided into a plurality of affine motion compensation image sub-blocks with the same size. Thus, the encoding complexity is reduced and the encoding performance is improved.

It should be understood that in this embodiment of the present invention, the processor 2010 may be a central processing unit (Central Processing Unit, "CPU" for short). Alternatively, processor 1910 may be another general purpose processor, digital signal processor (DSP), application specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete logic element or transistorized logic device, discrete hardware component or the like. A general purpose processor may be a microprocessor, or the processor may be any conventional processor or the like.

Memory 2030 may include read-only memory and random access memory, and provide instructions and data to processor 2010. Memory 2030 may further include non-volatile random access memory. For example, memory 2030 may additionally store device type information.

In addition to the data bus, the system bus 2050 may further include a power bus, a control bus, a status signal bus, and the like. However, for a clear description, the various types of busbars are marked as 2050 system bus in the drawing.

During implementation, the steps of the methods described above may be implemented by integrated logic circuitry hardware in the processor 2010, or by software instructions. The steps of the methods described with reference to embodiments of the present invention may be directly executed in a hardware processor, or may be performed using a combination of processor hardware and a software module. The program module may be located on a tangible storage medium known in the art, such as random access memory, flash memory, read-only memory, read-only programmable memory, electrically erasable programmable memory, or a register. The storage medium resides in memory 2030 and processor 2010 reads information from memory 2030 and performs the steps of the above methods in conjunction with the processor hardware. The repeated detailed description is omitted.

Claims (154)

1. A method for decoding a video image, comprising the steps of: determining the difference of the motion vector of the affine image block; determining the accuracy of the motion vector of the affine image block; determine the size of the affine sub-block of the motion compensation image in the affine image block based on the difference of the motion vector, the accuracy of the motion vector, the distance between the control points in the affine image block, the size of the affine image block, the size contains the length in the horizontal direction and the length in the vertical direction, and the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image subblock of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation in the vertical direction; and control points are pixels used to determine the motion vector difference; and performing decoding processing on the affine image block based on the size of the affine subblock of the motion compensation image. 2. The method of claim 1, wherein the step of determining the size of an affine image sub-block of a motion compensation image in the affine image block based on a motion vector difference, motion vector accuracy, a distance between control points in the affine image block, and a size of the affine image block, wherein the length of the affine image block is image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-block of the image motion compensation in the vertical direction, contains sub-steps, on which: proportional correction of the first horizontal distance is carried out based on the proportion of the motion vector accuracy to the first component of the motion vector difference to obtain the length of the affine motion compensation image subblock in the horizontal direction and the length of the affine subblock of the motion compensation image in the vertical direction, both lengths of the affine subblock of the motion compensation image in the horizontal direction direction and vertical direction are integers; determine whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer, a multiple of the length of the affine image sub-image sub-block of the motion compensation in the horizontal direction adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer, a multiple of the length of the affine image subblock of the image compensation in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock, so that the length of the vertical direction affine image block is an integer multiple of the adjusted length of the vertical motion compensation affine image subblock. 3. The method of claim 1, wherein the step of determining the size of an affine image sub-block of a motion compensation image in the affine image block based on a motion vector difference, motion vector accuracy, a distance between control points in the affine image block, and a size of the affine image block, wherein the length of the affine image block is image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-block of the image motion compensation in the vertical direction, contains sub-steps, on which: adjusting the first vertical distance proportionally based on the proportion of the motion vector accuracy to the second component of the motion vector difference to obtain the length of the motion compensation image affine sub-block in the vertical direction and the length of the motion compensation image affine sub-block in the horizontal direction, both lengths of the motion compensation image affine sub-block in the vertical direction and in the horizontal direction are integers; determine whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer, a multiple of the length of the affine image subblock of the image compensation in the vertical direction adjusting the length of the affine sub-image subblock of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image subblock of the vertical motion compensation image; and determine whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer, a multiple of the length of the affine image sub-image sub-block of the motion compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal motion compensation affine image subblock. 4. The method of claim 1, wherein the step of determining the size of an affine image subblock of the motion compensation image in the affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, wherein the length of the affine image block is image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-block of the image motion compensation in the vertical direction, contains sub-steps, on which: proportionally adjusting the first horizontal distance based on the proportion of the motion vector accuracy to the first motion vector difference component to obtain the length of the affine motion compensation image subblock in the horizontal direction, wherein the length of the affine subblock of the horizontal motion compensation image is an integer; determine whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer, a multiple of the length of the affine image sub-image sub-block of the motion compensation in the horizontal direction adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; proportionally adjusting the first vertical distance based on the proportion of the motion vector accuracy to the second motion vector difference component to obtain the length of the affine motion compensation image subblock in the vertical direction, wherein the length of the affine subblock of the vertical motion compensation image is an integer; and determining whether the vertical direction affine image block length is an integer multiple of the vertical direction motion compensation affine image subblock length, and if the vertical direction affine image block length is not an integer multiple of the vertical direction motion compensation affine image subblock length, adjusting the length of the affine image sub-block of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the adjusted length of the affine sub-image subblock of the vertical motion compensation image. 5. The method according to any one of claims 2-4, in which the step of adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock, comprising substeps of: adjusting, ie increasing/decreasing, the length of the affine sub-block of the motion compensation image in the horizontal direction by one or more unit lengths; and determine whether the length of the affine image block in the horizontal direction is an integer multiple of the adjusted length of the affine image block of the motion compensation image in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the horizontal direction, repeating the steps of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal motion compensation image; a the step of adjusting the length of the affine sub-image sub-block of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the adjusted length of the affine sub-image sub-block of the vertical motion compensation image, comprises sub-steps of: adjusting, that is, increasing/decreasing the length of the vertical motion compensation image affine sub-block by one or more unit lengths; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the direction, repeating the steps of adjusting and determining until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 6. The method according to any one of claims 2-4, in which the step of adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock, comprising substeps of: adjusting, that is, increasing/decreasing by one or more unit lengths, the length of the motion compensation affine image sub-block in the horizontal direction to the nearest value by which the length of the affine image block in the horizontal direction is divided; and determine whether the length of the affine image block in the horizontal direction is an integer multiple of the adjusted length of the affine image block of the motion compensation image in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the horizontal direction, repeating the steps of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion compensation; a the step of adjusting the length of the affine sub-image subblock of the vertical motion compensation image, wherein the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image subblock of the vertical motion compensation image, comprising sub-steps of: adjusting, i.e. increasing/decreasing by one or more unit lengths, the length of the affine sub-image compensation image block in the vertical direction to the nearest value by which the length of the affine image block in the vertical direction is divided; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the direction, repeating the steps of adjusting and determining until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 7. The method according to any one of claims 2 to 6, further comprising the steps of: determining whether the corrected length of the affine subblock of the horizontal motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the horizontal motion compensation affine image sub-block is not an integer multiple of the predetermined length S, re-correcting, that is, increasing/decreasing the corrected length of the horizontal motion compensation affine sub-image by one or more unit lengths, until the corrected length of the affine subblock of the horizontal motion compensation image will not be an integer multiple of the predetermined length S, in which S is equal to 2 ⁿ and n is zero or a positive number; and determining whether the corrected length of the affine sub-block of the vertical motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the vertical motion compensation image affine sub-block is not an integer multiple of the predetermined length S, re-correcting, that is, increasing/decreasing the corrected length of the vertical motion compensation image affine sub-block by one or more unit lengths, until the corrected length of the affine subblock of the vertical motion compensation image will not be an integer multiple of the predetermined length S, in which S is equal to 2 ⁿ and n is zero or a positive number. 8. The method of claim 1, wherein the step of determining the size of an affine subblock of a motion compensation image in the affine image block based on a motion vector difference, motion vector accuracy, a distance between control points in the affine image block, and a size of the affine image block, wherein the length of the affine image block is image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the vertical direction, contains a sub-step in which: determining, based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, the size of the motion compensation affine sub-image sub-block in the affine image block by looking at the predetermined size table, the predetermined size table storing a size representing motion compensation image affine subblock, and is predetermined based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, wherein the predetermined size of the affine image subblock of the motion compensation image corresponds to: the length of the affine image block in direction, which is an integer multiple of the length of the motion compensation affine image sub-block in the horizontal direction, and the length of the affine image block in the vertical direction, representing which is an integer multiple of the length of the affine subblock of the vertical motion compensation image. 9. The method of claim 8, wherein the predetermined size of the motion compensation affine image subblock further corresponds to: a length of the horizontal direction motion compensation affine image subblock that is an integer multiple of the predetermined length S, and a length of the vertical direction motion compensation affine image subblock , which is an integer multiple of the given length S, where S is 2 ⁿ and n is zero or a positive integer. 10. A method for encoding a video image, comprising the steps of: determining the difference of the motion vector of the affine image block; determining the accuracy of the motion vector of the affine image block; determine the size of the affine sub-block of the motion compensation image in the affine image block based on the difference of the motion vector, the accuracy of the motion vector, the distance between the control points in the affine image block and the size of the affine image block, wherein the size contains the length in the horizontal direction and the length in the vertical direction, and the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image subblock of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock of the motion compensation in the vertical direction; and control points, which are pixels used to determine the motion vector difference; and performing encoding processing on the affine image block based on the size of the affine subblock of the motion compensation image. 11. The method of claim 10, wherein the step of determining the size of an affine image sub-block of a motion compensation image in the affine image block based on a motion vector difference, motion vector precision, a distance between control points in the affine image block, and a size of the affine image block, wherein the length of the affine image block is image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-block of the image motion compensation in the vertical direction, contains sub-steps, on which: the first horizontal distance is proportionally corrected based on the proportion of the motion vector accuracy to the first component of the motion vector difference to obtain the length of the motion compensation image affine subblock in the horizontal direction and the length of the motion compensation image affine subblock in the vertical direction, both lengths of the motion compensation image affine subblock in horizontal direction and vertical direction are integers; determine whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer, a multiple of the length of the affine image sub-image sub-block of the motion compensation in the horizontal direction adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer, a multiple of the length of the affine image subblock of the image compensation in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock, so that the length of the vertical direction affine image block is an integer multiple of the adjusted length of the vertical motion compensation affine image subblock. 12. The method of claim 10, wherein the step of determining the size of an affine image subblock of a motion compensation image in the affine image block based on a motion vector difference, motion vector accuracy, a distance between control points in the affine image block, and a size of the affine image block, wherein the length of the affine image block is image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image sub-block of the image motion compensation in the vertical direction, contains sub-steps, on which: proportional adjustment of the first vertical distance based on the accuracy of the motion vector to the second component of the motion vector difference to obtain the length of the affine motion compensation image subblock in the vertical direction and the length of the affine subblock of the motion compensation image in the horizontal direction, both lengths of the affine subblock of the motion compensation image in the vertical direction and in the horizontal direction are integers; determine whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer, a multiple of the length of the affine image subblock of the image compensation in the vertical direction adjusting the length of the affine sub-image subblock of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image subblock of the vertical motion compensation image; and determine whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer, a multiple of the length of the affine image sub-image sub-block of the motion compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal motion compensation affine image subblock. 13. The method of claim 10, wherein the step of determining the size of an affine image subblock of the motion compensation image in the affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, wherein the length of the affine image block of the image block in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the horizontal direction motion compensation, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine sub-image of the compensation image in the vertical direction, contains sub-steps, on which : proportionally adjusting the first horizontal distance based on the proportion of the motion vector accuracy to the first motion vector difference component to obtain the length of the affine motion compensation image subblock in the horizontal direction, wherein the length of the affine subblock of the horizontal motion compensation image is an integer; determine whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer, a multiple of the length of the affine image sub-image sub-block of the motion compensation in the horizontal direction adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; proportionally adjusting the first vertical distance based on the proportion of the motion vector accuracy to the second motion vector difference component to obtain the length of the affine motion compensation image subblock in the vertical direction, wherein the length of the affine subblock of the vertical motion compensation image is an integer; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer, a multiple of the length of the affine image subblock of the image compensation in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock, so that the length of the vertical direction affine image block is an integer multiple of the adjusted length of the vertical motion compensation affine image subblock. 14. The method according to any one of claims 11-13, in which the step of adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock, comprising substeps of: adjusting, ie increasing/decreasing, the length of the affine sub-block of the motion compensation image in the horizontal direction by one or more unit lengths; and determine whether the length of the affine image block in the horizontal direction is an integer multiple of the adjusted length of the affine image block of the motion compensation image in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the the horizontal direction, repeating the step of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion correction; a the step of adjusting the length of the affine sub-image sub-block of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the adjusted length of the affine sub-image sub-block of the vertical motion compensation image, comprises sub-steps of: adjusting, that is, increasing/decreasing the length of the vertical motion compensation image affine sub-block by one or more unit lengths; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the direction, repeating the step of adjusting and determining until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 15. The method according to any one of claims 11-13, in which the step of adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock, comprising substeps of: adjusting, that is, increasing/decreasing by one or more unit lengths, the length of the motion compensation affine image sub-block in the horizontal direction to the nearest value by which the length of the affine image block in the horizontal direction is divided; and determine whether the length of the affine image block in the horizontal direction is an integer multiple of the adjusted length of the affine image block of the motion compensation image in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the horizontal direction, repeating the step of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion compensation; a the step of adjusting the length of the affine sub-image sub-block of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the adjusted length of the affine sub-image sub-block of the vertical motion compensation image, comprises sub-steps of: adjusting, i.e. increasing/decreasing by one or more unit lengths, the length of the affine sub-image compensation image block in the vertical direction to the nearest value by which the length of the affine image block in the vertical direction is divided; and determine whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the direction, repeating the step of adjusting and determining until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 16. A method according to any one of claims 11-15, further comprising the steps of: determining whether the corrected length of the affine subblock of the horizontal motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the horizontal motion compensation image affine sub-block is not an integer multiple of the predetermined length S, re-correcting, that is, increasing/decreasing the corrected length of the horizontal motion compensation image affine sub-block by one or more unit lengths up to those as long as the corrected length of the affine subblock of the horizontal motion compensation image is an integer multiple of the predetermined length S, where S is 2 ⁿ and n is zero or a positive number; and determining whether the corrected length of the affine sub-block of the vertical motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the vertical motion compensation image affine sub-block is not an integer multiple of the predetermined length S, re-correcting, that is, increasing/decreasing the corrected length of the vertical motion compensation image affine sub-block by one or more unit lengths, until the corrected length of the affine subblock of the vertical motion compensation image will not be an integer multiple of the predetermined length S, where S is equal to 2 ⁿ and n is zero or a positive number. 17. The method of claim 10, wherein the step of determining the size of the affine image sub-block of the motion compensation image in the affine image block based on the motion vector difference, the motion vector precision, the distance between control points in the affine image block, and the size of the affine image block, wherein the length of the affine image block is image in the horizontal direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine sub-image sub-block of the motion compensation in the vertical direction, contains sub-steps, on which: determining, based on the motion vector difference, the motion vector accuracy, the distances between control points in the affine image block, and the size of the affine image block, the size of the motion compensation affine sub-image sub-block in the affine image block by looking at a predetermined size table, the predetermined size table storing a size representing motion compensation image affine subblock and is set based on the motion vector difference, motion vector accuracy, distance between control points in the affine image block, and size of the affine image block, wherein the specified size of the affine subblocks of the motion compensation image satisfies the condition: length of the affine image block in the horizontal direction is an integer multiple of the length of the motion compensation affine image subblock in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of a layer that is a multiple of the length of the affine subblock of the vertical motion compensation image. 18. The method of claim 17, wherein the predetermined size of the motion compensation affine sub-image sub-frame is further satisfies: the length of the affine sub-image sub-frame of the motion compensation image in the horizontal direction is an integer multiple of the predetermined length S, and the length of the affine sub-image sub-frame of the motion compensation image in the vertical direction is is an integer multiple of the given length S, where S is 2 ⁿ and n is zero or a positive integer. 19. Video decoding equipment, comprising: a motion vector difference determination module, configured to determine a motion vector difference of the affine image block; a motion vector accuracy determination module, configured to determine the motion vector accuracy of the affine image block; an image sub-block size determination module, configured to determine the size of an affine image sub-block of a motion compensation image in the affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, wherein the size comprises a length in horizontal direction and length in the vertical direction, wherein the length of the affine image block in the horizontal direction is an integer multiple of the length of the motion compensation affine image subblock in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image subblock compensation for movement in the vertical direction; and reference points, which are pixels used to determine a motion vector difference; and a decoding module, configured to perform decoding on the affine image block based on the size of the affine sub-block of the motion compensation image. 20. The equipment of claim 19, wherein the image subblock size determination module is configured to: performing proportional adjustment of the first horizontal distance, based on the proportion of the motion vector accuracy to the first motion vector difference component, to obtain the length of the motion compensation affine sub-image in the horizontal direction and the length of the motion compensation affine sub-image in the vertical direction, both lengths of the motion compensation affine sub-image in the horizontal direction and vertical direction are integers; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; and determining whether the affine image block length in the vertical direction is an integer multiple of the length of the affine image sub-image block in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine image sub-image sub-block in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock so that the length of the vertical direction affine image block is an integer multiple of the corrected length of the vertical motion compensation affine image subblock. 21. The equipment of claim 19, wherein the image subblock size determination module is configured to: performing a proportional adjustment of the first vertical distance based on the proportion of the motion vector accuracy to the second component of the motion vector difference to obtain the length of the motion compensation affine sub-image in the vertical direction and the length of the affine sub-image of the motion compensation in the horizontal direction, both lengths of the motion compensation affine sub-image in the vertical direction and in the horizontal direction are integers; determining whether the affine image block length in the vertical direction is an integer multiple of the length of the affine image sub-image block in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine image sub-image sub-block in the vertical direction , adjusting the length of the affine image subblock of the vertical motion compensation image so that the length of the affine image block in the vertical direction is an integer multiple of the adjusted length of the affine image subblock of the vertical motion compensation image; and determining whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal motion compensation affine image subblock. 22. The equipment of claim 19, wherein the image subblock size determination module is configured to: proportionally adjusting the first horizontal distance based on the proportion of the motion vector accuracy to the first motion vector difference component to obtain the length of the affine motion compensation image subblock in the horizontal direction, wherein the length of the affine subblock of the horizontal motion compensation image is an integer; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; proportionally adjusting the first vertical distance based on the proportion of the motion vector accuracy to the second motion vector difference component to obtain the length of the vertical direction motion compensation image affine sub-block, wherein the length of the vertical direction motion compensation image affine sub-block is an integer; and determining whether the affine image block length in the vertical direction is an integer multiple of the length of the affine image sub-image block in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine image sub-image sub-block in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock so that the length of the vertical direction affine image block is an integer multiple of the adjusted length of the vertical motion compensation affine image subblock. 23. Equipment according to any one of claims 20-22, wherein the image sub-block size determination module is further configured to: adjusting, that is, increasing/decreasing the length of the affine subblock of the motion compensation image in the horizontal direction by one or more unit lengths; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in horizontal direction, repeating the steps of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion compensation; adjusting, that is, increasing/decreasing the length of the affine sub-block of the motion compensation image in the vertical direction by one or more unit lengths; and determining whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image block of the motion compensation image in the direction, repeating the steps of adjusting and determining until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 24. Equipment according to any one of claims 20-22, wherein the image sub-block size determination module is further configured to: adjusting, that is, increasing/decreasing by one or more unit lengths, the length of the motion compensation affine image subblock in the horizontal direction to the nearest value by which the length of the affine image block in the horizontal direction is divided; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in horizontal direction, repeating the steps of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion compensation; adjusting, that is, increasing/decreasing by one or more unit lengths, the length of the vertical motion compensation affine image subblock to the nearest value by which the length of the vertical direction affine image block is divisible; and determining whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the vertical direction direction, repeating the adjustment and determination steps until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 25. The method according to any one of claims 20-22, wherein the image sub-block size determination module is further configured to: determining whether the corrected length of the affine subblock of the horizontal motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the horizontal motion compensation affine sub-image sub-frame is not an integer multiple of the predetermined length S, re-correcting, that is, increasing/decreasing the corrected length of the horizontal motion compensation affine sub-frame by one or more unit lengths, until the re-corrected length of the affine subblock of the horizontal motion compensation image will not be an integer multiple of the predetermined length S, where S is 2 ⁿ and n is zero or a positive number; and determining whether the corrected length of the affine sub-block of the vertical motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the vertical motion compensation affine sub-image sub-frame is not an integer multiple of the predetermined length S, re-adjusting, that is, increasing/decreasing the corrected length of the vertical motion compensation affine sub-image by one or more unit lengths, while corrected the length of the affine subblock of the vertical motion compensation image will not be an integer multiple of the predetermined length S, where S is equal to 2 ⁿ and n is zero or a positive number. 26. The equipment of claim 19, wherein the image subblock size determination module is configured to: determining, based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block and the size of the affine image block, the size of the motion compensation affine sub-image sub-block in the affine image block, by looking at the predetermined size table, the predetermined size table storing the size, which is the motion compensation image affine sub-block, and is predetermined based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, and the predetermined size of the motion compensation affine image sub-block satisfies: the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine sub-image block of the motion compensation in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of length of the affine subblock of the vertical motion compensation image. 27. The equipment of claim 26, wherein the predetermined size of the motion compensation affine sub-image subframe further satisfies: the length of the affine sub-image sub-frame of the motion compensation image in the horizontal direction is an integer multiple of the predetermined length S, and the length of the affine sub-image sub-frame of the motion compensation image in the vertical direction is is an integer multiple of the given length S, where S is 2 ⁿ and n is zero or a positive integer. 28. Equipment for video encoding, comprising: a motion vector difference determination module, configured to determine a motion vector difference of the affine image block; a motion vector accuracy determination module, configured to determine the motion vector accuracy of the affine image block; an image sub-block size determination module configured to determine the size of an affine image sub-block of a motion compensation image in the affine image block based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, wherein the size comprises a length in in the horizontal direction and a length in the vertical direction, such that the length of the affine image block in the horizontal direction is an integer multiple of the length of the motion compensation affine image subblock in the horizontal direction, and the length of the affine image block in the vertical direction is an integer multiple of the length of the affine subblock images of motion compensation in the vertical direction; and control points being the pixels used to determine the motion vector difference; and an encoding unit configured to perform an encoding process on the affine image block based on the size of the affine sub-block of the motion compensation image. 29. The equipment of claim 28, wherein the image subblock size determination module is configured to: performing proportional adjustment of the first horizontal distance, based on the proportion of the motion vector accuracy to the first motion vector difference component, to obtain the length of the motion compensation affine sub-image in the horizontal direction and the length of the motion compensation affine sub-image in the vertical direction, both lengths of the motion compensation affine sub-image in the horizontal direction and vertical direction are integers; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; and determining whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine image subblock of the image compensation in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock so that the length of the vertical direction affine image block is an integer multiple of the adjusted length of the vertical motion compensation affine image subblock. 30. The equipment of claim 28, wherein the image subblock size determination module is configured to: proportional adjustment of the first vertical distance, based on the proportion of the motion vector accuracy to the second motion vector difference component, to obtain the length of the motion compensation affine sub-image in the vertical direction and the length of the motion compensation affine sub-image in the horizontal direction, both lengths of the motion compensation affine sub-image in vertical direction and horizontal direction are integers; determining whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine image subblock of the image compensation in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock so that the length of the vertical direction affine image block is an integer multiple of the adjusted length of the vertical motion compensation affine image subblock; and determining whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal motion compensation affine image subblock. 31. The equipment of claim 28, wherein the image subblock size determination module is configured to: proportionally adjusting the first horizontal distance based on the proportion of the motion vector accuracy to the first motion vector difference component to obtain the length of the affine motion compensation image sub-frame in the horizontal direction, wherein the length of the affine motion compensation image sub-frame in the horizontal direction is an integer; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the length of the affine image sub-image block in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the length of the affine image sub-image compensation in the horizontal direction , adjusting the length of the horizontal direction motion compensation affine image subblock so that the length of the horizontal direction affine image block is an integer multiple of the adjusted length of the horizontal direction motion compensation affine image subblock; proportionally adjusting the first vertical distance based on the proportion of the motion vector accuracy to the second motion vector difference component to obtain the length of the vertical motion compensation affine sub-image sub-frame, wherein the length of the vertical motion compensation affine sub-image is an integer; and determining whether the length of the affine image block in the vertical direction is an integer multiple of the length of the affine image block in the vertical direction motion compensation, and if the length of the affine image block in the vertical direction is not an integer multiple of the length of the affine image subblock of the image compensation in the vertical direction , adjusting the length of the vertical motion compensation affine image subblock so that the length of the vertical direction affine image block is an integer multiple of the adjusted length of the vertical motion compensation affine image subblock. 32. Equipment according to any one of claims 29-31, wherein the image sub-block size determination module is further configured to: adjusting, that is, increasing/decreasing the length of the affine subblock of the motion compensation image in the horizontal direction by one or more unit lengths; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in horizontal direction, repeating the steps of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion compensation; correcting, that is, increasing/decreasing the length of the affine sub-block of the motion compensation image in the vertical direction by one or more unit lengths; and determining whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image block of the motion compensation image in the direction, repeating the steps of adjusting and determining until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 33. Equipment according to any one of claims 29-31, wherein the image sub-block size determination module is further configured to: adjusting, that is, increasing/decreasing by one or more unit lengths, the length of the motion compensation affine image subblock in the horizontal direction to the nearest value by which the length of the affine image block in the horizontal direction is divisible; determining whether the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image sub-image of the motion compensation in the horizontal direction, and if the length of the affine image block in the horizontal direction is not an integer multiple of the corrected length of the affine sub-image of the motion compensation image in horizontal direction, repeating the steps of adjusting and determining until the length of the affine image block in the horizontal direction is an integer multiple of the corrected length of the affine sub-image block of the horizontal direction motion compensation; adjusting, that is, increasing/decreasing by one or more unit lengths, the length of the vertical motion compensation affine image subblock to the nearest value by which the length of the vertical direction affine image block is divisible; and determining whether the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine image block of the motion compensation image in the vertical direction, and if the length of the affine image block in the vertical direction is not an integer multiple of the adjusted length of the affine image subblock of the motion compensation image in the vertical direction direction, repeating the steps of adjusting and determining until the length of the affine image block in the vertical direction is an integer multiple of the corrected length of the affine sub-image block of the vertical motion compensation. 34. Equipment according to any one of claims 29-33, wherein the image sub-block size determination module is further configured to: determining whether the corrected length of the affine subblock of the horizontal motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the horizontal motion compensation affine sub-image sub-frame is not an integer multiple of the predetermined length S, re-correcting, that is, increasing/decreasing the corrected length of the horizontal motion compensation affine sub-frame by one or more unit lengths, while the corrected the length of the affine subblock of the horizontal motion compensation image will not be an integer multiple of the predetermined length S, where S is 2 ⁿ and n is zero or a positive number; and determining whether the corrected length of the affine sub-block of the vertical motion compensation image is an integer multiple of the predetermined length S, and performing, if the corrected length of the vertical motion compensation affine sub-image sub-block is not an integer multiple of the predetermined length S, re-correcting, that is, increasing/decreasing the corrected length of the vertical motion compensation affine sub-image by one or more unit lengths, while the corrected the length of the affine subblock of the vertical motion compensation image will not be an integer multiple of the given length S. 35. The equipment of claim 28, wherein the image subblock size determination module is configured to: determining, based on the motion vector difference, the motion vector accuracy, the distance between control points in the affine image block, and the size of the affine image block, the size of the motion compensation affine sub-image sub-block in the affine image block, by looking at the predetermined size table, the predetermined size table storing the size , which is an affine subblock of the motion compensation image and predetermined based on the difference of the motion vector, the accuracy of the motion vector, the distance between control points in the affine image block and the size of the affine image block, while the specified size of the affine subblock of the motion compensation image satisfies: the length of the affine image block in horizontal direction is an integer multiple of the length of the motion compensation affine image sub-block in the horizontal direction, and the length of the affine image block in the vertical direction is an integer th multiple of the length of the affine subblock of the vertical motion compensation image. 36. The equipment of claim 35, wherein the predetermined size of the motion compensation affine sub-image sub-unit further satisfies: the length of the affine sub-image sub-image of the motion compensation in the horizontal direction is an integer multiple of the predetermined length S, and the length of the affine sub-image sub-image of the motion compensation in the vertical direction is is an integer multiple of the given length S, where S is 2 ⁿ and n is zero or a positive integer. 37. A decoding device comprising a processor and read-only memory, wherein the processor is configured to call the program code stored in the memory to perform the decoding method according to any one of claims 1-9. 38. An encoding device comprising a processor and read-only memory, wherein the processor is configured to call the program code stored in the memory to implement the encoding method according to any one of claims 10-18.

Images