RU2701087C1 - Method and device for encoding and decoding motion vector based on reduced motion vector predictors-candidates - Google Patents

Abstract

FIELD: computer equipment.

SUBSTANCE: image decoding device comprises: decoding unit from bit stream information on motion vector difference for current block and information on motion vector predictor for current block; candidate determination unit; and motion vector recovery unit, wherein changed group of motion vector predictors includes one of first motion vector predictors and motion vector candidate second predictor, wherein first motion vector predictors are based on motion vectors of neighbouring blocks of current block, and second predictor-candidate motion vector is based on motion vector jointly located unit for current unit, and jointly located unit is in reference image, wherein the neighbouring units include a first unit located from below to the left of the current unit, and include a second unit located from the top of the first unit.

EFFECT: high compression ratio when decoding images.

9 cl, 28 dwg

Description

Technical field

[1] Devices and methods, in accordance with illustrative embodiments, relate to encoding and decoding a motion vector and, in particular, to predictive encoding and decoding of a motion vector of the current block.

State of the art

[2] In a codec such as "advanced video encoding" MPEG-4 H.264 / MPEG-4 (AVC), the motion vectors of previously encoded blocks adjacent to the current block can be used to predict the motion vector of the current block. The median of the motion vectors of previously encoded blocks adjacent to the left side, upper side and upper right side of the current block is used as a predictor of the motion vector of the current block. The motion vector of the current block is not directly encoded, but instead, the difference between the motion vector and the motion vector predictor is encoded.

Disclosure of invention

Solution

[3] One or more illustrative embodiments provides a method and apparatus for predictively encoding and decoding a motion vector, and a computer-readable recording medium on which a computer program for executing the method is recorded.

Advantageous Results of the Invention

[4] According to an illustrative embodiment, when motion vector candidate predictors are used to predictively encode and decode a motion vector, the number of motion vector predictor candidates can be reduced to predictively encode and decode a motion vector. Accordingly, the information necessary to indicate the motion vector predictor used to predict the motion vector of the current block from the motion vector predictor candidates can be encoded with a minimum number of bits, due to which the compression ratio of the encoding / decoding of the motion vector is increased, which allows to increase the compression ratio when encoding / image decoding.

Brief Description of the Drawings

[5] The above and other features and advantages will become apparent from the detailed description of illustrative embodiments with reference to the accompanying drawings, in which:

[6] FIG. 1 is a block diagram of an apparatus for encoding an image according to an illustrative embodiment;

[7] FIG. 2 is a block diagram of an apparatus for decoding an image according to an illustrative embodiment;

[8] FIG. 3 illustrates hierarchical coding units according to an illustrative embodiment;

[9] FIG. 4 is a block diagram of an image encoder based on a coding unit according to an illustrative embodiment;

[10] FIG. 5 is a block diagram of an image decoder based on a coding unit, according to an illustrative embodiment;

[11] FIG. 6 illustrates a maximum coding unit, a lower level coding unit, and a prediction unit, according to an illustrative embodiment;

[12] FIG. 7 illustrates a coding unit and a conversion unit, according to an illustrative embodiment;

[13] FIG. 8A to 8D illustrate separation forms of a coding unit, a prediction unit, and a conversion unit, according to an illustrative embodiment;

[14] FIG. 9 is a block diagram of an apparatus for encoding a motion vector according to an illustrative embodiment;

[15] FIG. 10A and 10B illustrate motion vector predictor candidates according to an illustrative embodiment;

[16] FIG. 10C-10E illustrate blocks of various sizes that are adjacent to the current block, according to an illustrative embodiment;

[17] FIG. 11A-11C illustrate motion vector predictor candidates according to another illustrative embodiment;

[18] FIG. 12 illustrates a method for reducing the number of motion vector predictor candidates according to an illustrative embodiment;

[19] FIG. 13A-13D illustrate the location of a current block included in a coding unit of a predetermined size, according to an illustrative embodiment;

[20] FIG. 14 is a block diagram of an apparatus for decoding a motion vector according to an illustrative embodiment;

[21] FIG. 15 is a flowchart illustrating a method of encoding a motion vector according to an illustrative embodiment; and

[22] FIG. 16 is a flowchart illustrating a method for decoding a motion vector according to an illustrative embodiment.

Preferred Embodiments

[23] According to an aspect of an illustrative embodiment, a method for encoding a motion vector is provided, the method comprising the steps of: generating information about the motion vector based on the motion vector of the current block and the motion vector predictor of the current block by estimating the motion vector of the current block and determining a first motion vector predictor candidate from a plurality of motion vector predictor candidates as a motion vector predictor of the current block based on ultat assessment; generating a virtual motion vector using a second motion vector predictor candidate from a plurality of motion vector predictor candidates and motion vector information, generating vector differences between a virtual motion vector and a plurality of motion vector predictors, comparing the vector differences with motion vector information, and selectively excluding a second motion vector predictor candidate from a plurality of motion vector predictor candidates; and encode motion vector information and motion vector predictor information of the current block.

[24] According to an aspect of another illustrative embodiment, a method for decoding a motion vector is provided, the method including the steps of: decoding information about a motion vector of a current block; generating a virtual motion vector using a predetermined motion vector predictor candidate from a plurality of motion vector predictor candidates and decoded motion vector information, generating vector differences between a virtual motion vector and a plurality of motion vector predictors, comparing the generated vector differences with decoded information about motion vector, and selectively exclude a predetermined predictor candidate motion vector from a set of motion vector predictor candidates; and determining the motion vector predictor of the motion vector predictor candidates that are not excluded from the plurality of motion vector predictors as the motion vector predictor of the current block, and restoring the motion vector of the current block based on the determined motion vector predictor and decoded motion vector information.

[25] According to an aspect of another illustrative embodiment, there is provided an apparatus for encoding a motion vector, the apparatus including: a motion vector estimator that generates motion vector information based on a motion vector of the current block and a motion vector predictor of the current block by estimating the motion vector the current block and determines the first predictor candidate motion vector from the set of predictor candidates of the motion vector as a predictor of the vector d izheniya current block based on the estimation result; a candidate determination unit that generates a virtual motion vector using a second motion vector predictor candidate from a plurality of motion vector predictor candidates and motion vector information generates vector differences between a virtual motion vector and a plurality of motion vector predictors, compares vector differences with information about the motion vector, and selectively excludes the second predictor candidate motion vector from the set of predictor candidates vector motion; and a motion vector encoder that encodes motion vector information and motion vector predictor information of the current block.

[26] According to an aspect of another illustrative embodiment, an apparatus for decoding a motion vector is provided, the apparatus including: a motion vector decoder that decodes information about a motion vector of a current block; a candidate determination unit that generates a virtual motion vector using a predetermined motion vector predictor candidate from a plurality of motion vector predictor candidates and decoded motion vector information generates vector differences between a virtual motion vector and a plurality of motion vector predictors, compares the generated vector differences with decoded motion vector information, and selectively excludes a predetermined prediction spruce candidate motion vector from the set of predictor candidates motion vector; and a motion vector recovery unit that determines a motion vector predictor candidate of the motion vector predictor candidates that are not excluded from the plurality of motion vector predictor candidates as a motion vector predictor of the current block and restores a motion vector of the current block based on the determined motion vector predictor and decoded motion vector information.

[27] According to an aspect of another illustrative embodiment, a computer-readable recording medium is provided that contains a computer program for performing motion vector encoding and decoding methods.

Embodiments of the invention

[28] Next, one or more illustrative embodiments will be described more fully with reference to the accompanying drawings. Expressions such as “at least one of” preceding the list of items change the list of items as a whole and do not change individual items in the list.

[29] Further, the 'image' may mean a still image for a video or a moving image, that is, the video itself.

[30] in FIG. 1 is a block diagram of an image encoding apparatus 100 according to an exemplary embodiment.

[31] Referring to FIG. 1, an image encoding apparatus 100 includes a maximum coding unit dividing unit 110, an encoding depth determining unit 120, an image data encoder 130, and an encoding information encoder 140.

[32] The maximum coding unit division unit 110 may divide the current frame or slice based on the maximum coding unit, that is, the largest coding unit. Thus, the maximum coding unit division unit 110 may divide the current frame or slice into at least one maximum coding unit.

[33] According to an illustrative embodiment, a coding unit may be represented using a maximum coding unit and depth. As described above, the maximum coding unit indicates the coding unit having the largest size of the coding units of the current frame, and the depth indicates the degree of hierarchical reduction of the coding unit. As the depth increases, the coding unit may decrease from the maximum coding unit to the minimum coding unit, wherein the depth of the maximum coding unit is defined as the minimum depth, and the depth of the minimum coding unit is determined as the maximum depth. Since the size of the coding unit decreases from the maximum coding unit with increasing depth, the coding unit of a lower level of the kth depth may include many coding units of a lower level of the (k + n) th depth (k and n are integers greater than or equal to 1).

[34] According to the increase in the size of the frame to be encoded, encoding an image in a larger encoding unit may result in a higher image compression ratio. However, if a larger coding unit is fixed, efficient coding of an image reflecting continuously changing image characteristics may not be possible.

[35] For example, when encoding a smooth area, such as the sea or sky, the larger the coding unit, the more the compression ratio can be increased. However, when encoding a complex area, such as people or buildings, the smaller the coding unit, the more the compression ratio can be increased.

[36] Accordingly, according to an illustrative embodiment, for each frame or slice, a different maximum coding unit and a different maximum depth can be set. Since the maximum depth indicates the maximum number of times that a coding unit can be reduced, the size of each minimum coding unit included in the maximum coding unit can be set differently according to the maximum depth.

[37] Block 120 determine the encoding depth determines the maximum depth. The maximum depth can be determined on the basis of the calculation of the cost of "speed-distortion" (RD). The maximum depth can be defined differently for each frame or slice, or for each maximum coding unit. The determined maximum depth is supplied to the encoding information encoder 140, and image data according to the maximum encoding units is supplied to the image data encoder 130.

[38] The maximum depth may indicate the smallest encoding unit that can be included in the maximum encoding unit, that is, the minimum encoding unit. In other words, the maximum coding unit may be divided into lower level coding units having different sizes according to different depths, which will be described below with reference to FIG. 8A and 8B. In addition, lower level coding units having different sizes that are included in the maximum coding unit can be predicted or converted based on processing units having different sizes. The conversion is performed to convert the pixel values of the spatial domain to the coefficients of the frequency domain and may be a discrete cosine transform or Karunen-Loeve transform (KLT). In other words, the image encoding apparatus 100 may perform a variety of processing operations for encoding an image based on processing units having different sizes and different shapes. To encode image data, processing operations such as at least one of prediction, conversion, and statistical coding are performed, and processing units having the same size or different sizes can be used for each operation.

[39] For example, the device 100 for encoding an image may select a processing unit that is different from a coding unit for predicting a coding unit.

[40] When the size of the coding unit is 2N × 2N (where N is a positive integer), the processing units for prediction may be 2N × 2N, 2N × N, N × 2N, and N × N. In other words, motion prediction can be based on a processing unit having a shape such that at least one of the height and width of the coding unit is divided into two equal parts. Hereinafter, the processing unit, which is the basis of the prediction, is defined as ‘prediction unit’.

[41] The prediction mode can be at least one of an internal mode, a mutual mode, and a skip mode, and a specific prediction mode can be performed only for a prediction unit having a specific size or specific shape. For example, the internal mode can be carried out only for the unit of prediction, sizes 2N × 2N and N × N and square. In addition, the skip mode can only be performed for a prediction unit of size 2N × 2N. If there are many prediction units in the coding unit, then after the prediction is completed, for each prediction unit, the prediction mode with the least coding errors can be selected.

[42] Alternatively, the image encoding apparatus 100 may convert image data based on a processing unit different in size from the encoding unit. For conversion in a coding unit, the conversion may be based on a processing unit whose size is less than or equal to the size of the coding unit. Hereinafter, the processing unit, which is the basis of the conversion, is defined as ‘conversion unit’.

[43] The coding depth determination unit 120 may determine lower level coding units included in the maximum coding unit using RD optimization based on the Lagrange multiplier. In other words, the coding depth determining unit 120 may determine the shape of multiple lower-level coding units obtained by dividing the maximum coding unit, the many lower-level coding units having different sizes according to their depths. The image data encoder 130 outputs a bit stream by encoding a maximum coding unit based on the separation forms determined by the encoding depth determining unit 120.

[44] The coding information encoder 140 encodes coding mode information of a maximum coding unit determined by coding depth determining unit 120. In other words, the encoding information encoder 140 outputs a bit stream encoding information on a separation form of a maximum encoding unit, information on a maximum depth, and information on an encoding mode of a lower level encoding unit for each depth. Information about the encoding mode of the lower coding unit may include information about the prediction unit of the lower coding unit, information about the prediction mode for each prediction unit, and information about the conversion unit of the lower coding unit.

[45] Information about the separation form of the maximum coding unit may be information indicating whether each coding unit is divided. For example, when the maximum coding unit is divided and encoded, information indicating whether the maximum coding unit is divided is encoded. Furthermore, when a lower level coding unit obtained by dividing the maximum coding unit is divided and encoded, information indicating whether the lower level coding unit is divided is encoded. Information indicating whether a lower level coding unit is divided may be flag information.

[46] Since lower level coding units having different sizes exist for each maximum coding unit, and coding mode information needs to be determined for each lower coding unit, information about at least one coding mode can be determined for one maximum unit coding.

[47] The image encoding apparatus 100 may generate lower level coding units by dividing in two equal parts the height and width of the maximum coding unit according to the increase in depth. Thus, when the size of the coding unit of the kth depth is 2N × 2N, the size of the coding unit of the (k + 1) depth is N × N.

[48] Accordingly, the image encoding apparatus 100 may determine an optimal separation form for each maximum coding unit based on the sizes of the maximum coding units and the maximum depth, taking into account image characteristics. By flexibly adjusting the size of the maximum coding unit taking into account the characteristics of the image, and encoding the image by dividing the maximum coding unit into coding units of a lower level of different depths, it is possible to more effectively encode images having different resolutions.

[49] In FIG. 2 is a block diagram of an apparatus 200 for decoding an image according to an illustrative embodiment.

[50] Referring to FIG. 2, an image decoding apparatus 200 includes an image data acquiring unit 210, encoding information extraction unit 220, and an image data decoder 230.

[51] The image data acquisition unit 210 obtains image data according to the maximum coding units by analyzing the bit stream received by the image decoding apparatus 200 and outputs the image data to the image data decoder 230. The image data acquiring unit 210 may extract information about the maximum coding unit of the current frame or slice from the header of the current frame or slice. In other words, the image data acquiring unit 210 divides the bitstream into a maximum coding unit, which allows the image data decoder 230 to decode the image data according to the maximum coding units.

[52] The coding information extraction unit 220 extracts information about a maximum coding unit, a maximum depth, a separation form of a maximum coding unit, a coding unit encoding mode of a lower level from the header of the current frame, analyzing the bitstream received by the image decoding apparatus 200. Information about the form of separation and information about the encoding mode is supplied to the decoder 230 of the image data.

[53] Information on the separation form of the maximum coding unit may include information on lower level coding units having different sizes according to depths and included in the maximum coding unit, and may be information (eg, flag information) indicating whether each coding unit. The coding mode information may include prediction unit information according to coding units of a lower layer, prediction mode information, and transform unit information.

[54] The image data decoder 230 restores the current frame by decoding the image data of each maximum coding unit based on information extracted by the coding information extraction unit 220.

[55] The image data decoder 230 may decode lower-level coding units included in the maximum coding unit based on the split form information of the maximum coding unit. The decoding process may include a mutual prediction process including intra prediction and motion compensation, and an inverse transformation process.

[56] The image data decoder 230 may perform intra prediction or inter prediction based on prediction unit information according to coding units of a lower layer and prediction mode information for predicting a coding unit of a lower layer. The image data decoder 230 may also perform the inverse transform for each coding unit of a lower level based on information about a conversion unit of a coding unit of a lower level.

[57] FIG. 3 illustrates hierarchical coding units according to an illustrative embodiment.

[58] Referring to FIG. 3, hierarchical coding units may include coding units for which the width × height is 64 × 64, 32 × 32, 16 × 16, 8 × 8, and 4 × 4. In addition to these coding units having the correct square shape, coding units may also exist for which the width × height is 64 × 32, 32 × 64, 32 × 16, 16 × 32, 16 × 8, 8 × 16, 8 × 4, and 4 × 8.

[59] In FIG. 3, for image data 310 with a resolution of 1920 × 1080, the size of the maximum coding unit is set to 64 × 64, and the maximum depth is set to 2.

[60] For image data 320 with a resolution of 1920 × 1080, the size of the maximum encoding unit is set to 64 × 64, and the maximum depth is set to 3. For data 330 images with a resolution of 352 × 288, the size of the maximum encoding unit is set to 16 × 16, and the maximum depth is set to 2.

[61] At high resolution or a large amount of data, the maximum size of a coding unit can be relatively large to increase the compression ratio and accurately reflect image characteristics. Accordingly, for image data 310 and 320 having a higher resolution than image data 330, 64 × 64 can be selected as the size of the maximum coding unit.

[62] The maximum depth indicates the total number of levels in hierarchical coding units. Since the maximum depth of the image data 310 is 2, the encoding unit 315 of the image data 310 may include a maximum encoding unit whose size along the long axis is 64, and lower level encoding units whose sizes along the long axis are 32 and 16, according to the increase depths.

[63] On the other hand, since the maximum depth of the image data 330 is 2, the encoding unit 335 of the image data 330 may include a maximum encoding unit whose size along the long axis is 16 and encoding units whose dimensions along the long axis are 8 and 4, according to the increase in depth.

[64] However, since the maximum depth of image data 320 is 4, the encoding unit 325 of image data 320 may include a maximum encoding unit whose size along the long axis is 64, and lower level encoding units whose dimensions along the long axis are 32 , 16, 8 and 4, according to the increase in depth. Since the image is encoded based on a lower coding unit of a lower level with increasing depth, an illustrative embodiment is suitable for encoding an image including more small scenes.

[65] In FIG. 4 shows a block diagram of an image encoder 400 based on a coding unit, according to an illustrative embodiment.

[66] The intra prediction unit 410 performs intra prediction on the intra mode prediction units in the current frame 405, and the motion estimation unit 420 and the motion compensation unit 425 perform mutual prediction and motion compensation on the mutual mode prediction units using the current frame 405 and the reference frame 495 .

[67] The residual values are generated based on the prediction units output from the intra prediction unit 410, the motion estimation unit 420 and the motion compensation unit 425, and the generated residual values are output as quantized transform coefficients by passing through the converter 430 and the quantization unit 440.

[68] The quantized transform coefficients are restored to residual values by passing through inverse quantization block 460 and inverse transformer 470, and the reconstructed residual values are further processed by passing through block elimination unit 480 and loop filtering unit 490 and output as a reference frame 495 The quantized transform coefficients may be output as bitstream 455 by passing through a statistical encoder 450.

[69] To encode based on the encoding method according to an illustrative embodiment, the components of the image encoder 400, i.e. intra prediction block 410, motion estimation block 420, motion compensation block 425, transducer 430, quantization block 440, statistic encoder 450, inverse quantization block 460, inverse transducer 470, block division elimination block and contour filtering block 490, coding processes are carried out images based on the maximum coding unit, lower level coding unit according to depths, prediction units and conversion units.

[70] In FIG. 5 shows a block diagram of an image decoder 500 based on a coding unit, according to an illustrative embodiment.

[71] Referring to FIG. 5, bitstream 505 passes through an analyzer 510, which allows decoding encoded image data and analyzing encoding information used for decoding. The encoded image data is output as inverse-quantized data by passing through a statistical decoder 520 and an inverse quantization unit 530 and reconstructed to residual values by passing through an inverse converter 540. Residual values are restored according to the coding units by summing the internal prediction unit 550 of the intra prediction or the result of the motion compensation unit 560 motion compensation. Recovered coding units are used to predict the next coding units or the next frame by passing block separation elimination block 570 and loop filtering block 580.

[72] For decoding based on a decoding method according to an illustrative embodiment, the components of the image decoder 500, i.e. an analyzer 510, a statistical decoder 520, an inverse quantization unit 530, an inverse transducer 540, an intra prediction unit 550, a motion compensation unit 560, a block division elimination unit 570, and a loop filtering unit 580, image decoding processes are performed based on the maximum coding unit, coding unit lower level according to depths, prediction units and conversion units.

[73] In particular, the intra prediction unit 550 and the motion compensation unit 560 determine the prediction unit and the prediction mode in the lower level coding unit based on the maximum coding unit and depth, and the inverter 540 performs the inverse transformation based on the size of the transformation unit.

[74] FIG. 6 illustrates a maximum coding unit, a lower level coding unit, and a prediction unit, according to an illustrative embodiment.

[75] The image encoding apparatus 100 illustrated in FIG. 1, and the image decoding apparatus 200 illustrated in FIG. 2, hierarchical coding units are used to encode and decode based on image characteristics. The maximum coding unit and maximum depth can be adaptively set according to image characteristics or variable set according to user requirements.

[76] In FIG. 6, the hierarchical coding unit structure 600 has a maximum coding unit 610, the height and width of which is 64, and a maximum depth of 4. The depth increases along the vertical axis of the hierarchical coding unit structure 600, and as the depth increases, the height values of the width of units 620 to 650 lower level encodings are reduced. Prediction units of the maximum coding unit 610 and lower coding units 620 to 650 are shown along the horizontal axis of the hierarchical coding unit structure 600.

[77] The maximum coding unit 610 has a depth of 0 and the size of the coding unit, i.e. height and width, 64 × 64. The depth increases along the vertical axis, and there are a lower level encoding unit 620 of size 32 × 32 and a depth of 1, a lower level encoding unit 630 of size 16 × 16 and a depth of 2, a lower level encoding unit 630 of size 8x8 and a depth of 3, and a unit 650 encoding a lower level of 4 × 4 and a depth of 4. Unit 650 encoding a lower level of 4 × 4 and a depth of 4 is the minimum coding unit.

[78] Referring to FIG. 6, examples of a prediction unit are shown along a horizontal axis according to each depth. Thus, the prediction unit of the maximum coding unit 610, the depth of which is 0, may be a prediction unit of the same size as the coding unit 610, i.e. 64 × 64, or a 64 × 32 prediction unit 612, a 32 × 64 prediction unit 614, or a 32 × 32 prediction unit 616 that is smaller than the 64 × 64 encoding unit 610.

[79] The prediction unit of the coding unit 620, the depth of which is 1 and the size is 32 × 32, may be a prediction unit of the same size as the coding unit 620, i.e. 32 × 32, or a 32 × 16 prediction unit 622, a 16 × 32 prediction unit 624, or a 16 × 16 prediction unit 626 that is smaller than the 32 × 32 encoding unit 620.

[80] A prediction unit of a coding unit 630 whose depth is 2 and a size of 16 × 16 may be a prediction unit of the same size as the coding unit 630, ie 16 × 16, or a 16 × 8 prediction unit 632, an 8 × 16 prediction unit 634, or an 8 × 8 prediction unit 636 that is smaller than the 16 × 16 encoding unit 630.

[81] A prediction unit of a coding unit 640, the depth of which is 3 and a size of 8 × 8, may be a prediction unit of the same size as the coding unit 640, i.e. 8 × 8, or an 8 × 4 prediction unit 642, a 4 × 8 prediction unit 644, or a 4 × 4 prediction unit 646 smaller than that of an 8 × 8 encoding unit 640.

[82] Finally, a coding unit 650, the depth of which is 4 and a size of 4 × 4, is a maximum depth coding unit, and the prediction unit of a coding unit 650 may be a 4 × 4 prediction unit 650. However, the maximum depth coding unit may not always be the same size as the prediction unit. Similar to coding units 610 to 650, the maximum depth coding unit may be divided into prediction units having smaller dimensions than the coding unit, and may predict.

[83] FIG. 7 illustrates a coding unit and a conversion unit, according to an illustrative embodiment.

[84] The image encoding apparatus 100 illustrated in FIG. 1, and the image decoding apparatus 200 illustrated in FIG. 2, encoding and decoding is performed with the maximum coding unit or lower level coding units that are less than or equal to the maximum coding unit obtained by dividing the maximum coding unit. In the encoding and decoding process, the size of the transform unit for the transform is selected so as to provide the highest compression ratio regardless of the encoding unit or prediction unit. For example, as shown in FIG. 7, when the current coding unit 710 is 64 × 64 in size, the conversion can be performed using a 32 × 32 conversion unit 720.

[85] FIG. 8A through 8D illustrate separation forms of a coding unit 810, a prediction unit 860, and a conversion unit 870, according to an illustrative embodiment.

[86] FIG. 8A and 8B illustrate a coding unit 810 and a prediction unit 860, according to an illustrative embodiment.

[87] In FIG. 8A shows a separation form selected by apparatus 100 for encoding an image illustrated in FIG. 1, for encoding a maximum coding unit 810. The image encoding apparatus 100 divides the maximum encoding unit 810 into various forms, performs encoding and selects an optimal separation form by comparing the encoding results of the various separation forms with each other based on the costs of the RD. When optimally encoding the maximum coding unit 810 as is, the maximum coding unit 810 can be encoded without dividing the maximum coding unit 810 as illustrated in FIG. 8A - 8D.

[88] Referring to FIG. 8B, the maximum coding unit 810, the depth of which is 0, is encoded by dividing it into lower coding units whose depths are greater than or equal to 1. Thus, the maximum coding unit 810 is divided into 4 coding units of a lower depth level of 1, and all or some of the coding units of a lower level whose depths are 1 are divided into coding units of a lower level whose depths are 2.

[89] The lower-level coding unit located on the upper right side and the lower-level coding unit located on the lower left side, from lower-level coding units whose depths are 1, are divided into lower-level coding units whose depths greater than or equal to 2. Some of the lower level encoding units whose depths are greater than or equal to 2 can be divided into lower level encoding units whose depths are greater than or equal to 3.

[90] In FIG. 8B shows a separation form of a prediction unit 860 for a maximum coding unit 810.

[91] Referring to FIG. 8B, the prediction unit 860 for the maximum coding unit 810 may be divided differently than the maximum coding unit 810. In other words, the prediction unit for each of the lower level coding units may be less than the corresponding lower level coding unit.

[92] For example, a prediction unit for a lower layer encoding unit 854 located on the lower right side of among lower level encoding units whose depths are 1 may be less than a lower level encoding unit 854. In addition, the prediction units for some lower-level coding units 814, 816, 850, and 852 of the lower-level coding units 814, 816, 818, 828, 850, and 852, whose depths are 2, may be less than 814, 816, 850 and 852 coding lower level, respectively.

[93] Furthermore, prediction units for lower level coding units 822, 832 and 848, whose depths are 3, may be less than lower level coding units 822, 832 and 848, respectively. Prediction units may take the form in which the corresponding lower-level coding units are divided into two equal parts in the direction of height or width, or in the form in which the corresponding lower-level coding units are divided into four equal parts in the direction of height and width.

[94] FIG. 8C and 8D illustrate a prediction unit 860 and a transform unit 870, according to an illustrative embodiment.

[95] In FIG. 8C shows a separation form of the prediction unit 860 for the maximum coding unit 810 shown in FIG. 8B, and in FIG. 8D shows a separation form of a conversion unit 870 of a maximum coding unit 810.

[96] Referring to FIG. 8D, the division form of the transform unit 870 can be set differently than for the prediction unit 860.

[97] For example, although the prediction unit for the coding unit 854, the depth of which is 1, is selected with a form in which the height of the coding unit 854 is divided into two equal parts, the conversion unit can be selected the same size as the coding unit 854. Similarly, although prediction units for coding units 814 and 850, the depths of which are 2, are selected with a form in which the height of each of the coding units 814 and 850 is divided into two equal parts, the conversion unit can be selected the same size as the original size p of each of the units 814 and 850 coding.

[98] A conversion unit can be selected with a smaller size than a prediction unit. For example, when a prediction unit for a coding unit 852 whose depth is 2 is selected with a shape in which the width of the coding unit 852 is divided into two equal parts, a transform unit can be selected with a shape in which the coding unit 852 is divided in the height and width directions by four equal parts that are smaller than the shape of the prediction unit.

[99] In FIG. 9 is a block diagram of an apparatus 900 for encoding a motion vector according to an illustrative embodiment.

[100] The apparatus 900 for encoding a motion vector may be included in the apparatus 100 for encoding the image illustrated in FIG. 1, or the image encoder 400 illustrated in FIG. 4 is shown in detail in FIG. 9. According to FIG. 9, a motion vector encoding apparatus 900 includes a motion vector estimator 910, a candidate determination unit 920, and a motion vector encoder 930.

[101] For decoding encoded blocks using mutual prediction, ie temporal prediction, information is required on the motion vector, which indicates the relative difference of locations between the current block and a similar block in the reference image. Accordingly, information about the motion vector is encoded with simultaneous encoding of the image and embedded in the bitstream. With the direct encoding and implementation of information about the motion vector, the overhead used to encode information about the motion vector increases, and thus, the compression ratio of the image data is reduced.

[102] Accordingly, when encoding an image, the motion vector of the current block is predicted, and only the difference of the initial motion vector between the motion vector predictor generated by the prediction and the original motion vector is subjected to encoding and embedding, which allows compressing information about the motion vector.

[103] In predictive coding of a motion vector, an explicit mode and an implicit mode may exist.

[104] In a codec such as Advanced Video Coding MPEG-4 H.264 / MPEG-4 (AVC), motion vectors of previously encoded blocks adjacent to the current block can be used to predict the motion vector of the current block. The median of the motion vectors of previously encoded blocks adjacent to the left side, upper side, and upper right side of the current block is used as a predictor of the motion vector of the current block. Since the motion vectors of all blocks encoded using mutual prediction are predicted using the same method, information about the motion vector predictor of the current block may not be encoded separately. However, the image encoding apparatus 100 or image encoder 400 according to one or more illustrative embodiments uses both an implicit mode in which motion vector predictor information is not encoded separately and an explicit mode in which motion vector predictor information is not encoded for accurate prediction of the motion vector. In explicit mode, information about the motion vector predictor used as the motion vector predictor of the current block from the set of motion vector predictor candidates is encoded and embedded into the bitstream as a sequence parameter, slice parameter, or block parameter.

[105] FIG. 9 illustrates a device that performs predictive coding, while a motion vector is encoded according to an explicit mode.

[106] The motion vector estimator 910 estimates the motion vector of the current block. A block similar or identical to the current block is searched in at least one reference image, and, as a result of the search, a motion vector is estimated, which is the relative position difference between the current block and the desired reference image. A block similar or identical to the current block is searched based on the calculation of the absolute difference sum (SAD), and, as a result of the search, the motion vector of the current block can be estimated.

[107] In addition, the motion vector estimator 910 predicts the motion vector of the current block based on the motion vectors of the blocks included in a previously encoded region adjacent to the current block. In other words, the motion vectors of blocks included in a previously encoded region adjacent to the current block are set as motion vector predictors, and a motion vector predictor is determined that is most similar to the estimated motion vector of the current block from motion vector predictor candidates.

[108] In a codec such as MPEG-4 H.264 / MPEG-4 AVC, the median of the motion vectors of previously encoded blocks adjacent to the left side, upper side, and upper right side of the current block is used as a predictor of the motion vector of the current block. Since the motion vector of the encoded block is predicted using the motion vectors of the previously encoded blocks, and only one motion vector predictor is used, information about the motion vector predictor may not be separately encoded. In other words, the number of predictors of the motion vector of a block encoded using mutual prediction is one.

[109] However, when the motion vector of the current block is accurately predicted, the motion vector can be encoded with a high compression ratio. In this regard, according to an illustrative embodiment, one of the plurality of motion vector predictor candidates is selected and used as a motion vector predictor of the current block, which allows encoding a motion vector of the current block with a high compression ratio. Next, a method for encoding a motion vector of a current block using a plurality of motion vector predictor candidates will be described in detail.

[110] FIG. 10A and 10B illustrate motion vector predictor candidates according to an illustrative embodiment.

[111] Referring to FIG. 10A, according to the motion vector prediction method, one of the motion vectors of previously encoded blocks adjacent to the current block can be used as a predictor of the motion vector of the current block.

[112] Of the blocks adjacent to the upper side of the current block, all the motion vectors of block a0 are on the very left side, block b0 is on the highest side, adjacent to the left side, block c is adjacent to the upper right side, block d is adjacent to the left the upper side, and block e adjacent to the lower left side can be used as predictor candidates for the motion vector of the current block.

[113] According to the method of encoding and decoding an image according to one or more illustrative embodiments, the encoding and decoding of the image is carried out on the basis of coding units having different sizes and different depths, and thus a motion vector of a block adjacent to the lower left side, can be used as a candidate for motion vector predictors.

[114] Referring to FIG. 10B, the motion vectors of all blocks adjacent to the current block can be used as motion vector predictor candidates. In other words, the motion vectors not only of the block a0 on the left side of the blocks adjacent to the upper side, but also all blocks a0 - aN adjacent to the upper side, can be used as motion vector predictor candidates. In addition, the motion vectors of not only the block b0 on the uppermost side of the blocks adjacent to the left side, but also all the blocks b0 - bN adjacent to the left side, can be used as motion vector predictor candidates.

[115] In addition, the median of the motion vectors of neighboring blocks can be used as predictor candidates for the motion vector. In other words, the median mv_a0, mv_b0 or mv_c can be used as a candidate for the predictors of the motion vector of the current block. Here, mv_a0 is the motion vector of block a0, mv_b0 is the motion vector of block b0, and mv_c is the motion vector of block c.

[116] However, candidate motion vector predictors of the current block may be limited according to the sizes of the current block and neighboring blocks, which will be described in detail with reference to FIG. 10C-10E.

[117] FIG. 10C-10E illustrate blocks of various sizes that are adjacent to the current block, according to an illustrative embodiment.

[118] As described above, according to the method of encoding and decoding an image according to one or more illustrative embodiments, coding units and a prediction unit having different sizes determined according to depths can be used to encode the image. Accordingly, since the blocks adjacent to the current block may have different sizes, when the size of the current block and the sizes of some neighboring blocks are significantly different from each other, the motion vectors of some neighboring blocks may not be used as motion vector predictor candidates.

[119] Referring to FIG. 10C, blocks 1014-1018 adjacent to the upper side of the current block 1010 are smaller than the current block 1010. Since the motion vector of block 1012 adjacent to the current block 1010 and having the same size as the current block 1010 may be identical or similar to the motion vector of the current block 1010, motion vector estimator 910 can only use the motion vector of block 1012 as a candidate for motion vector predictors.

[120] Although the sizes are not the same, only motion vectors of neighboring blocks having predetermined sizes or more can be used as predictor candidates for the motion vector. For example, only motion vectors of blocks 1012 and 1018, the sizes of which are 1/4 of the size of the current block 1010 or more, can be used as predictors of motion vector candidates.

[121] Referring to FIG. 10D, the size of the block 1022 adjacent to the left side of the current block 1020 exceeds the size of the current block 1020 by 16 times, and therefore, there is a big difference in size. Due to this large difference, the motion vector of block 1022 adjacent to the left side may not be identical or similar to the motion vector of current block 1020. Accordingly, the motion vector of block 1022 adjacent to the left side is not used as a predictor candidate motion vector of the current block 1020, and as motion vector predictor candidates, only the motion vectors of block 1024 adjacent to the upper side and block 1026 adjacent to the left upper side can be used.

[122] Referring to FIG. 10E, the size of the current block 1030 exceeds the sizes of all neighboring blocks 1031-1037. Here, when the motion vectors of all neighboring blocks 1031-1037 are used as predictor candidates for the motion vector of the current block 1030, the number of predictor candidates for the motion vector of the current block 1030 can be large. As the size difference between the current block 1030 and the neighboring blocks 1031-1037 increases, the number of motion vector predictor candidates also increases. Accordingly, the motion vector estimator 910 may not use the motion vectors of some blocks from neighboring blocks as predictor candidates for the motion vector of the current block 1030.

[123] For example, in FIG. 10E, the motion vectors of block 1031 adjacent to the lower left side and block 1037 adjacent to the upper right side may not be used as motion predictors of the motion vector of current block 1030.

[124] In a more general case, when the size of the current block 1030 exceeds a predetermined size, the motion vectors of blocks neighboring in a particular direction from neighboring blocks may not be used as motion predictors of the motion vector of the current block 1030.

[125] FIG. 11A-11C illustrate motion vector predictor candidates according to another illustrative embodiment.

[126] FIG. 11A illustrates a method for determining a predictor candidate of a two-way image motion vector (image B). When the current image including the current block is an image B that performs two-way prediction, the motion vector generated based on the time distance may be a motion vector predictor candidate.

[127] The predictor candidate of the motion vector mv_temporal of the current block 1100 of the current image 1110 can be determined using the motion vector of the block 1120 located next to the current block 1100 of the image 1112, which precedes the current image 1110 in time. For example, when the motion vector mv_colA of the block 1120 is next to the current block 1100, is generated with respect to the desired block 1122 of the image 1114, which follows the time after the current image 1110, mv_L0A and mv_L1A, which are predictors of the motion vector candidates The current block 1100 can be defined as follows:

[128] mv_L1A = (t1 / t2) * mv_colA

[129] mv_L0A = mv_L1A-mv_colA

[130] Here, mv_L0A denotes a predictor candidate motion vector of the current block 1110 with respect to the image 1112, which precedes the time of the current image 1110, and mv_L1A denotes the predictor candidate motion vector of the current block 1100 with respect to the image 1114 that follows the time after the current image 1110.

[131] In FIG. 11A, the current image 1110, which is the image B, is inserted between the image 1112, which precedes the time of the current image 1110, and the image 1114, which follows the time after the current image 1110. Here, when the motion vector mv_colA of the block 1120 located next to the current block 1100, is generated with respect to the image 1114, which follows the time after the current image 1110, the motion vector of the current block 1100 can be accurately predicted based on mv_L1A. In other words, the motion vector of the current block 1100 can be more accurately predicted when mv_colA is a motion vector in the direction illustrated in FIG. 11 A, as compared to the case where mv_colA is a motion vector whose direction is opposite to that illustrated in FIG. 11 A, that is, when mv_colA is generated with respect to another image prior to the image 1112, which is preceded in time by the current image 1110.

[132] Accordingly, when the direction from the current block 1110 to the block 1120 adjacent to the current block 1100 is List0, the motion vector mv_colA should be oriented in the direction List1, and thus, the current image 1110 can be inserted between the image 1112 and image 1114, as illustrated in FIG. 11A, which allows accurate prediction of the motion vector of the current block 1100 based on mv_colA.

[133] In addition, images 1110 to 1114 illustrated in FIG. 11A are presented according to a temporal order, and thus, the motion vector predictor predictor mv_temporal of the current block can be generated based on the image sequence number (POC). The image referenced by the current block may be an image different from the neighboring images 1112 and 1114 illustrated in FIG. 11A, and thus the motion vector predictor predictor of the current block is generated based on the POC.

[134] For example, when the POC of the current image is CurrPOC, and the POC of the image referenced by the current image is CurrRefPOC, a motion vector predictor predictor of the current block can be generated as follows:

[135] Scale = (CurrPOC-CurrRefPOC) / (ColPOC-ColRefPOC)

[136] mv_temporal = Scale * mv_colA

[137] Here, ColPOC is a POC of an image 1112 that precedes the current image 1110 and includes a block 1120 and ColRefPOC is a POC of an image 1114 that follows the time after the current image 1110 and includes a block 1122, referenced by block 1120.

[138] FIG. 11B illustrates another method for generating a motion vector predictor predictor of image B. The method illustrated in FIG. 11B is different from the method shown in FIG. 11A in that the image 1114 that follows the time after the current image 1110 includes a block adjacent to the current block 1100.

[139] Referring to FIG. 11B, the motion vector predictor candidate of the current block 1100 of the current image 1110 may be generated using the motion vector of the block 1130 located next to the current block 1100 of the image 1114, which follows the time after the current image 1110. For example, when the motion vector mv_colB of the block 1130 located next to the current block 1100, is generated with respect to the desired block 1132 of the image 1112, which is preceded in time by the current image 1110, mv_L0B and mv_L1B, which are predictors of the motion vector candidates the current block 1100, can be determined as follows:

[140] mv_L0B = (t3 / t4) * mv_colB

[141] mv_L1B = mv_L0B-mv_colB

[142] Here, mv_L0B denotes the predictor candidate motion vector of the current block 1110 with respect to the image 1112, which precedes the time of the current image 1110, and mv_L1B denotes the predictor candidate motion vector of the current block 1100 with respect to the image 1114 that follows the time after the current image 1110.

[143] Similarly to FIG. 11A, in FIG. 11B, the current image 1110, which is the image B, is inserted between the image 1112, which precedes the time of the current image 1110, and the image 1114, which follows the time after the current image 1110. Accordingly, when the motion vector mv_colB of the block 1130 located next to the current block 1100, is generated with respect to the image 1112, which is preceded in time by the current image 1110, the motion vector of the current block 1100 can be accurately predicted based on mv_L0B. In other words, the motion vector of the current block 1100 can be more accurately predicted when mv_colB is a motion vector in the direction illustrated in FIG. 11B, compared with the case where mv_colB is a motion vector whose direction is opposite to that illustrated in FIG. 11B, that is, when mv_colB is generated with respect to another image after the image 1114, which follows the time after the current image 1110.

[144] Accordingly, when the direction from the current block 1110 to the block 1130 adjacent to the current block 1100 is List1, the motion vector mv_colB of the block 1130 should be oriented in the direction List0, and thus, the current image 1110 can be inserted between the image 1112 and image 1114, as illustrated in FIG. 11B, which allows accurate prediction of the motion vector of the current block 1100 based on mv_colB.

[145] Furthermore, the image referenced by the current unit may be an image different from the neighboring images 1112 and 1114 illustrated in FIG. 11B, and thus the motion vector predictor predictor of the current block is generated based on the POC.

[146] For example, when the POC of the current image is CurrPOC, and the POC of the image referenced by the current image is CurrRefPOC, a motion vector predictor predictor of the current block can be generated as follows:

[147] Scale = (CurrPOC-CurrRefPOC) / (ColPOC-ColRefPOC)

[148] mv_temporal = Scale * mv_colB

[149] Here, ColPOC is a POC of an image 1114 that follows the time after the current image 1110 and includes a block 1130, and ColRefPOC is a POC of an image 1112 that precedes the time of the current image 1110 and includes a block 1132, the link to which provided by block 1130.

[150] FIG. 11C illustrates a candidate predictor of a motion vector of a predictive image (image P).

[151] Referring to FIG. 11C, the motion vector predictor candidate of the current block 1100 of the current image 1110 can be determined using the motion vector of the block 1140 located next to the current block 1100 of the image 1112, which is preceded in time by the current image 1110. For example, when the motion vector mv_colC of the block 1140 is adjacent to the current block 1100, is generated with respect to the desired block 1142 of the image 1116, which is preceded in time by the current image 1110, mv_L0C, which is a predictor candidate motion vector of the current block ka 1100 can be defined as follows:

[152] mv_L0C = (t6 / t5) * mv_colC

[153] As described in connection with FIG. 11A and 11B, mv_L0C may be determined based on the POC. mv_L0C can be determined based on the POC of the current image 1110, the POC of the image referenced by the current image 1110, the POC of the image 1112 that precedes the time of the current image 1110, and the POC of the image 1116 that precedes the time of the current image 1110.

[154] Since the current image 1110 is an image P, only one motion vector predictor predictor of the current block 1100 is determined, in contrast to FIG. 11A and 11B.

[155] In addition, to use the motion vector predictor predictor generated based on the time distance in FIG. 11A and 11B for predicting a motion vector of a current block, it is also possible to encode information indicating which block of blocks 1120 and 1130 adjacent to the current block 1100 is used to generate a motion vector predictor. Such information may be included in the slice header or sequence header as a slice parameter or sequence parameter.

[156] In general, the motion vector predictor group C of FIG. 10A, 10B and 11A-11C may have the following form:

[157] C = {median (mv_a0, mv_b0, mv_c), mv_a0, mv_a1, ..., mv_aN, mv_b0, mv_b1, ..., mv_bN, mv_c, mv_d, mv_e, mv_temporal}

[158] In addition, group C may be a group with a reduced number of motion vector predictor candidates:

[159] C = {median (mv_a ', mv_b', mv_c '), mv_a', mv_b ', mv_c', mv_temporal}

[160] Here, mv_x denotes the motion vector of block x, median () denotes an average value, and mv_temporal denotes motion vector predictor candidates generated using the time distance described in connection with FIG. 11A-11C. mv_a 'denotes the first effective motion vector of mv_a0, mv_a1, ..., and mv_aN. For example, when block a0 is encoded using intra prediction and indicates the current block and another image, the motion vector mv_a0 of block a0 is not effective, and thus mv_a '= mv_a1. In addition, when the motion vector of block a1 is not effective, mv_a '= mv_a2. Similarly, mv_b 'is the first effective motion vector of mv_b0, mv_b1, ..., and mv_bN and mv_c' is the first effective motion vector of mv_c, mv_d and mv_e.

[161] A motion vector of a block that indicates an image different from the current block from the motion vectors of blocks adjacent to the current block may inefficiently predict the motion vector of the current block. Accordingly, a motion vector of a block that indicates an image different from the current block can be excluded from the group C of motion vector predictor candidates.

[162] When the apparatus 900 for encoding a motion vector encodes a motion vector according to the explicit mode, the apparatus 900 also encodes information indicating which motion vector predictor candidate in group C is used to predict the motion vector of the current block. In other words, when the motion vector encoding apparatus 900 encodes a motion vector, the apparatus 900 assigns binary numbers to the corresponding elements of group C, i.e., motion vector predictors, and encodes binary numbers corresponding to the motion vector predictor candidate used to predict the motion vector of the current block.

[163] To indicate one of the elements in group C, binary numbers are assigned and displayed, each of which corresponds to the predictors of the motion vector candidates. Thus, as the number of elements in group C decreases, elements in group C can be indicated with binary numbers with fewer bits.

[164] Accordingly, in the presence of overlapping predictor candidates of the motion vector in group C, binary numbers can be assigned after excluding the overlapping predictor candidates of the motion vector from group C. For example, when the group C is C = {median (mv_a ', mv_b', mv_c '), mv_a', mv_b ', mv_c', mv_temporal}, as described above, and mv_a ', mv_b' and mv_c 'coincide with each other, it can be determined that there are two elements in group C, namely, C = {mv_a ', mv_temporal} and binary numbers may be assigned. When five elements in group C can be indicated using 3 bits before excluding overlapping motion vector predictors, two elements can be indicated using 1 bit after eliminating overlapping motion vector predictors.

[165] To increase the likelihood of determining an overlapping motion vector predictor candidate as a motion vector predictor of the current block, instead of eliminating the overlapping motion vector predictor candidate, a predetermined weight can be assigned to it. As described above, since mv_a ', mv_b' and mv_c 'are the same and only mv_a' is included in group C, mv_a 'is assigned a predetermined weight, and thus the probability that mv_a' is used to predict the motion vector of the current block may increase.

[166] Furthermore, if there is one motion vector predictor candidate, the binary number used to indicate one of the motion vector predictor candidates may not be encoded. For example, when the group C is C = {median (mv_a0, mv_b0, mv_c), mv_a0, mv_a1, ..., mv_aN, mv_b0, mv_b1, ..., mv_bN, mv_c, mv_d, mv_e, mv_temporal}, and blocks a0-aN, and b0-bN, block c, block d, and block e are all subject to mutual prediction, group C is C = {mv_temporal} and, thus, essentially includes one element. Accordingly, in this case, the apparatus 900 for encoding a motion vector may not encode a binary number used to indicate one of the motion vector predictor candidates.

[167] It will be apparent to those skilled in the art that motion vectors other than the motion vector predictor candidates described above can be used as motion vector predictor candidates.

[168] Furthermore, according to another exemplary embodiment, the candidate determination unit 920 may reduce the number of motion vector predictor candidates based on the evaluation of the encoding result.

[169] As described above, individual information is encoded and included in the bitstream to indicate a motion vector predictor candidate used to predict a motion vector of a current block from a plurality of motion vector predictor candidates. Accordingly, as the number of elements in group C decreases, the information necessary to indicate the motion vector predictor candidate used to predict the motion vector of the current block in group C can be encoded with a small number of bits. In this regard, the candidate determination unit 920 may selectively exclude predetermined motion vector predictors from all motion vector predictors using a predetermined evaluation function, which will be described in detail with reference to FIG. 12.

[170] FIG. 12 illustrates a method for reducing the number of motion vector predictor candidates according to an illustrative embodiment.

[171] In FIG. 12, it is assumed that there are three elements MVP1, MVP2, and MVP3 in group C, and the motion vector of the current block is denoted by MV. Since the motion vector predictor candidate that is most similar to the motion vector of the current block is used to predict the motion vector of the current block, MVP3, which is most similar to MV, is used to predict the motion vector of the current block.

[172] Accordingly, the vector difference (hereinafter referred to as первонач initial motion vector difference ’) between the motion vector of the current block and the predictor motion vector candidate used to predict the motion vector of the current block is (2.0). Since MV is (5.0) and MVP3 is (3.0), the difference of the initial motion vector is (2.0).

[173] The candidate determination unit 920 selectively excludes at least one motion vector predictor candidate from all motion vector predictor candidates using a difference of the initial motion vector and a predetermined evaluation function. In particular, the difference of the initial motion vector and the predetermined predictor-candidate motion vector are used to generate a virtual motion vector, and the differences (hereinafter referred to as the "differences of the virtual motion vector") between the generated virtual motion vector and all predictors of the motion vector candidates are generated with respect to all to candidates. The difference of the initial motion vector and the predetermined predictor-candidate motion vector are summed with each other to generate a virtual motion vector, and the differences of the virtual motion vector between the generated virtual motion vector and all predictors of the motion vector candidates are calculated. By comparing the difference of the initial motion vector with the differences of the virtual motion vector calculated with respect to all candidates, a predetermined predictor candidate motion vector can be selectively excluded from all predictor candidates of the motion vector.

[174] Referring to FIG. 12, the candidate determination unit 920 determines whether MVP1, one of the motion vector predictor candidates, is excluded from the total number of candidates.

[175] When the difference of the virtual motion vector generated by subtracting the motion predictor candidate of the motion vector from the virtual motion vector based on MVP1 is less than the difference of the initial motion vector, MVP1 can not be used to predict the motion vector of the current block. For example, when the difference of the virtual motion vector generated by subtracting MVP3 from the virtual motion vector generated by summing MVP1 and the difference of the initial motion vector is less than the difference of the initial motion vector, MVP3 predicts the virtual motion vector more accurately than MVP1, in which case MVP1 cannot be motion vector predictor.

[176] In FIG. 12, when MVP1 and the difference of the initial motion vector are added together, the virtual motion vector based on MVP1 is (2.0). Accordingly, when the virtual motion vector is generated based on MVP1, the difference of the virtual motion vector for MVP2 is (2.0), and the difference of the virtual motion vector for MVP3 is (-1.0). Here, since the difference of the virtual motion vector (-1.0) for MVP3 is less than the difference of the initial motion vector (2.0), MVP1 may not be a predictor of the motion vector of the current block. Accordingly, MVP1 can be excluded from all motion vector predictor candidates. In other words, the motion vector predictor candidate that corresponds to MVP1 can be excluded from group C.

[177] Here, in itself, the difference of the virtual motion vector calculated for MVP1 is (2.0) and is always identical to the difference of the initial motion vector, so the difference of the virtual motion vector can be no less than the difference of the initial motion vector. Accordingly, when the differences of the virtual motion vector are calculated for all predictors-candidates of the motion vector, the difference of the virtual motion vector for MVP1 per se can not be calculated.

[178] Upon completion of determining whether MVP1 is excluded, candidate determination unit 920 determines whether MVP2 is excluded from all motion vector predictor candidates. When MVP2 and the difference of the initial motion vector are summed with each other, the virtual motion vector based on MVP2 is (2.0). Accordingly, the difference of the virtual motion vector for MVP1 is (2.0), and the difference of the virtual motion vector for MVP3 is (-1.0). Since the difference of the virtual motion vector for MVP3 is less than the difference of the initial motion vector, MVP2 can be excluded from all predictor candidates of the motion vector, as in the case of MVP1. When determining whether MVP2 is excluded, the difference of the virtual motion vector for MVP1 can be selectively compared with the difference of the initial motion vector. Since it has already been determined that MVP1 should be excluded, the differences of the virtual motion vector for candidates other than MVP1 can be compared with the difference of the initial motion vector.

[179] In addition, the candidate determination unit 920 determines whether MVP3 is excluded. The virtual motion vector based on MVP3 is identical to the original motion vector of the current block. Although another motion vector predictor candidate (i.e., MVP1 or MVP2) is subtracted from the original motion vector, a virtual motion vector difference that is less than the difference of the original motion vector may not be generated. Thus, MVP3 is not excluded from the motion vector predictor candidates. In addition, according to another illustrative embodiment, since MVP3, according to the determination result to be used to predict the motion vector of the current block, is not excluded from the motion vector predictors, the candidate determination unit 920 may skip determining whether MVP3, the motion vector predictor candidate, used to predict the motion vector of the current block.

[180] In short, when the candidate determination unit 920 determines whether a second motion vector predictor candidate, one of the motion vector predictor candidates, a second motion vector predictor candidate, and a difference of the initial motion vector are added together to generate a virtual motion vector , and the differences between the virtual motion vector and other predictors, the motion vector candidates are calculated for all candidates to generate the set of differences of the virtual motion vector izheniya. When there is at least one difference of the virtual motion vector, which is less than the difference of the initial motion vector, from the set of differences of the virtual motion vector, the second predictor candidate motion vector is not a predictor motion vector of the current block and is excluded from the predictor candidates of the motion vector.

[181] In addition, the candidate determination unit 920 repeats the determination of an exception for the motion vector predictor candidates, which reduces the number of motion vector predictor candidates, that is, elements of group C. According to the order of placement of the predictor candidates of the motion vector of group C, the need exceptions are determined in order. For example, when C = {median (mv_a ', mv_b', mv_c '), mv_a', mv_b ', mv_c', mv_temporal}, the need to exclude median (mv_a ', mv_b', mv_ ') is determined, and upon completion of the determination, the need to exclude mv_a 'is determined. Then, the need to exclude mv_b 'is determined. According to the C grouping order, the definition of the exception is repeated until mv_temporal.

[182] When the determination is carried out in a repeated manner, the comparison between the difference of the virtual motion vector and the difference of the initial motion vector for candidates already excluded in the determination can be omitted, as described in connection with the exception of MVP2.

[183] In addition, group C can be reordered according to a predetermined criterion, which will be described below with reference to FIG. 13A-13D. When group C is reordered, the definition of the exception is repeated according to the reordered order.

[184] The comparison between the difference of the virtual motion vector and the difference of the initial motion vector described in connection with FIG. 12, can be applied not only to the one-dimensional motion vector, but also to the two-dimensional motion vector. In other words, the absolute value of the difference of the virtual motion vector specified by x-coordinates and y-coordinates is compared with the absolute value of the difference of the initial motion vector, and thus, a predetermined motion vector predictor can be selectively excluded from the total number of candidates.

[185] However, the absolute value used to compare the difference of the virtual motion vector with the difference of the initial motion vector is given only as an example, and various criteria can be used to compare the difference of the virtual motion vector with the difference of the initial motion vector. When the evaluation function that generates the value for the difference of the virtual motion vector and the value for the difference of the initial motion vector based on a predetermined criterion is ‘A’, the difference of the virtual motion vector can be compared with the difference of the initial motion vector according to the following equation 1:

[186] (equation 1)

[187] A (mvx + MVD-mvy) <A (MVD)

[188] The candidate determination unit 920 determines whether at least one 'mvy' exists in all candidates to determine whether 'mvx', one of the motion vector predictor candidates, is excluded from the motion vector candidate predictor. In equation 1, 'MVD' denotes the difference of the initial motion vector. To determine the need to exclude 'mvx', 'A (mvx + MVD-mvy)', which is the value obtained by evaluating 'mvx + MVD-mvy', i.e. the difference of the virtual motion vector between 'mvx + MVD', the virtual motion vector based on 'mvx' and 'mvy', another motion vector candidate predictor, is calculated using the predetermined evaluation function 'A', and the values generated by the calculation are compared with 'A (MVD)', the value for the difference of the original vector movement. Motion vector predictor candidates other than the 'mvx' of all candidates are repeatedly substituted for 'mvy', and it is determined whether there is at least one mvy in the total number of candidates that satisfies equation 1.

[189] As described above, the difference of the virtual motion vector and the difference of the initial motion vector estimated according to ‘A’ can be specified by x-coordinates and y-coordinates. In this case, the estimated function may be defined by the sum of the estimated x-coordinate value and the estimated y-coordinate value, according to the following equation 2:

[190] (equation 2)

[191] A (p, q) = f (p) + f (q)

[192] When the difference of the virtual motion vector or the difference of the initial motion vector is specified by the x-coordinates 'p' and y-coordinates 'q', each coordinate value is entered into a predetermined function 'f', and the estimated function 'A' is given by the sum obtained through substitution.

[193] According to an illustrative embodiment, the estimated function ‘A’ in Equation 1 and Equation 2 may be an estimated function that estimates the result obtained by statistical coding of the difference of the virtual motion vector and the result obtained by statistical coding of the difference of the original motion vector. The candidate determination unit 920 estimates the result obtained by statistical coding the difference of the virtual motion vector and the difference of the initial motion vector based on the evaluation function ‘A’, and the number of motion vector predictor candidates can be reduced based on the evaluation result. This is described in detail with reference to the following equation 3:

[194] (equation 3)

[195] Length = 1;

[196] Temp = (val <= 0)? (-val << 1) +1: (val << 1);

[197] while (1! = Temp) {

[198] Temp >> = 1;

[199] Length + = 2;

[200]}

[201] f (val) = Length

[202] The function 'f', which evaluates the result obtained by statistical coding with respect to the x-coordinate value or y-coordinate values, can be specified according to equation 3. When the 'val', the x-coordinate or y-coordinate value, is entered to the function 'f', which evaluates variable length encoding, the result (for example, universal variable length encoding), 'Length' is calculated according to equation 3.

[203] Equation 3 can be represented as follows:

[204]

[205] The x-coordinate or y-coordinate value may be the x-coordinate or y-coordinate value of the difference of the virtual motion vector or the difference of the initial motion vector.

[206] According to equation 3, when 'val' is a negative number or ‘0’, 'val' becomes a positive number and then multiplied by ‘2’ by shifting left by 1 bit. And ‘1’ is added, which allows you to save the result in 'Temp'. When 'val' is a positive number, 'val' is multiplied by ‘2’ by shifting left by 1 bit, which allows you to save the result in 'Temp'.

Then the 'while' loops are repeated until the 'Temp' is equal to 1, and the 'Length' is computed.

[207] For example, when the difference of the virtual motion vector or the difference of the initial motion vector is (2.0), A (2.0) = f (2) + f (0).

[208] f (2) is calculated as follows. ‘2’ f (2) is a positive number and thus shifts to the left by 1 bit, so that 'Temp' is set to ‘4’. In the first loop, 'while' 'Temp' is equal to '4', that is, not equal to '1', and thus, '4' is multiplied by '1/2' by shifting to the right, as a result of which 'Temp' is set to '2 '. Since the initial value of 'Length' is set to ‘1’, 'Length' in the first cycle of 'while' is equal to '3'.

[209] In the second loop, 'while' 'Temp' is equal to '2', that is, not equal to '1', and thus '2' is multiplied by '1/2' by shifting to the right, as a result of which 'Temp' is set equal to 'one'. Since the current value of 'Length' is 3, in the second loop, 'while' 'Length' is '5'. The third cycle of 'while' is not implemented, since 'Temp' is ‘1’, and f (2) is '5'.

[210] f (0) is calculated as follows. Since the input value of the coordinate f (0) is ‘0’, '0' is shifted left by 1 bit, and ‘1’ is added, so that 'Temp' is set to ‘1’. Accordingly, while loops are not implemented. f (0) is ‘1’ according to the initial value of 'Length'.

[211] The predefined evaluation function ‘f’ described in connection with Equation 3 is a function for evaluating the result of statistical coding using variable length coding. Accordingly, the candidate determination unit 920 evaluates the result obtained by encoding the variable length of the differences of the virtual motion vector using the evaluation function ‘A’ to determine whether 'mvx' is excluded from the motion vector predictor candidates. As a result of the evaluation, when at least one difference of the virtual motion vector exists, according to the estimate, to be encoded to a shorter length than the difference of the initial motion vector, mvx is excluded from all motion vector predictor candidates.

[212] However, it is obvious to a person skilled in the art that the result of statistical coding is evaluated using methods other than the result of coding of variable length. For example, to evaluate and compare the result of statistical coding of the difference of the virtual motion vector and the result of statistical coding of the difference of the initial motion vector, you can use another evaluation function ‘h’. Here, ‘h’ may be a function that evaluates the result of context adaptive binary arithmetic coding.

[213] Furthermore, according to another exemplary embodiment, to improve the accuracy of the evaluation result based on a predetermined evaluation function, the result obtained by evaluating the index information can also be evaluated. Index information is used to indicate a predetermined motion vector predictor candidate from all motion vector predictor candidates. This is described in detail with reference to equation 4:

[214] (equation 4)

[215] A (mvx + MVD-mvy, mvyIdx) <A (MVD, mvxIdx)

[216] The candidate determination unit 920 determines whether there is at least one 'mvy' that satisfies Equation 4 in the total number of candidates to determine whether 'mvx', one of the motion vector predictor candidates, is excluded from the vector predictor candidates movement. In Equation 4, 'MVD' stands for the difference of the initial motion vector, and mvxIdx and mvyIdx denote the index information used to indicate ‘mvx’ and 'mvy', respectively, of all motion vector predictor candidates. To determine the need to exclude 'mvx', 'mvx + MVD-mvy', i.e. the difference of the virtual motion vector between 'mvx + MVD', the virtual motion vector based on 'mvx', and 'mvy', another predictor of the motion vector candidate, and index information for indicating 'mvy' of all candidates is evaluated using a predetermined evaluation function 'A'. In addition, the difference of the initial motion vector and the index information for indicating ‘mvx’ from the total number of candidates is estimated using the predetermined evaluation function ‘A’. As a result of the evaluation, it is determined whether at least one 'mvy' exists in the total number of candidates.

[217] As described above, the difference of the virtual motion vector and the difference of the initial motion vector estimated according to ‘A’ can be given by x-coordinates and y-coordinates and can be set according to the following equation 5:

[218] (equation 5)

[219] A (mvx + MVD-mvy, mvyIdx) = f (p1) + f (q1) + g (mvyIdx)

[220] A (MVD, mvxIdx) = f (p2) + f (q2) + g (mvxIdx)

[221] Compared to Equation 2, A (mvx + MVD-mvy, mvyIdx) of Equation 5 on the left side of Equation 2 estimates only the difference of the virtual motion vector, while A (mvx + MVD-mvy, mvyIdx) of Equation 5 estimates the difference of the virtual motion vector and information to indicate 'mvy' of all motion vector predictor candidates. The evaluation function ‘A’ may be a function for evaluating the result of statistical coding. Here, the function 'f' may be a function for evaluating the result of statistical coding based on the x-coordinate value or y-coordinate value of the difference of the virtual motion vector, as described in connection with equation 2, and the function 'g' may be a function for evaluating the result of the statistical encoding 'mvxIdx'. When the x-coordinate value and the y-coordinate value of 'mvx + MVD-mvy' are equal to 'p1' and 'q1', respectively, A (mvx + MVD-mvy, mvxIdx) can be calculated according to equation 5.

[222] A (MVD) on the right side of equation 2 estimates only the difference of the initial motion vector, while A (MVD, mvxIdx) of equation 5 estimates the difference of the initial motion vector and information to indicate 'mvx' from all motion vector predictor candidates. The function 'f' may be a function for evaluating the result of statistical coding based on the x-coordinate value or the y-coordinate of the difference of the initial motion vector, as described in connection with equation 2, and the function 'g' may be a function for evaluating the result of statistical encoding ' mvxIdx '. When the x-coordinate value and the y-coordinate value of 'MVD' are equal to 'p2' and 'q2', respectively, A (MVD, mvxIdx) can be calculated according to equation 5.

[223] The definition of an exception according to equations 4 and 5 can be additionally used to determine the exception according to equation 2. In other words, whether 'mvx' is excluded from all motion vector predictor candidates is first determined based on equations 2, and the exception can be determined again according to equations 4 and 5. For example, as a result of the determination according to Equation 2, when 'A (mvx + MVD-mvy)' is greater than or equal to 'A (MVD)' and 'A (mvx + MVD-mvy)' is not less than 'A ( MVD) ',' mvx 'is not excluded from all motion vector predictor candidates according to avneniyu 2. However, although 'A (mvx + MVD-mvy)' and 'A (MVD)' equal, 'mvx' can be eliminated from all predictors candidate motion vector on basis of a determination result according to equations 4 and 5.

[224] When the candidate determination unit 920 determines an exception for the motion vector predictor candidates based on equations 1-5, the determination of the exception for the motion vector predictor candidates is repeated in accordance with the arrangement of group C. According to another illustrative embodiment, the candidate determination unit 920 reorders the group C according to a predetermined criterion, and the determination of the exception can be repeated according to the reordered order. This is described in detail with reference to FIG. 13A-13D.

[225] FIG. 13A-13D illustrate the location of a current block included in a coding unit of a predetermined size, according to an illustrative embodiment.

[226] When all the motion vector predictor candidates form C = {median (mv_a ', mv_b', mv_c '), mv_a', mv_b ', mv_c', mv_temporal}, each of the binary numbers is assigned to the predictors as candidates for the motion vector of group C, and thus, the motion vector predictor candidate used to predict the motion vector of the current block from the motion vector predictor candidates can be indicated, as described above.

[227] Here, binary numbers are assigned according to the order of placement of motion vector predictor candidates included in group C, and may be a variable-length code based on a Huffman code. Accordingly, a small number of bits can be assigned to motion vector predictor candidates located in front in the order of group C. For example, bit '0' can be assigned to 'median (mv_a', mv_b ', mv_c') ', bit' 00 'can be assigned to mv_a' , and bit '01' may be assigned to mv_b '. Accordingly, the candidate determination unit 920 organizes motion vector predictor candidates according to a predetermined order, whereby the motion vector predictor used with high probability to predict the motion vector of the current block of motion vector predictor candidates is located in front of group C.

[228] The motion vector predictor used with high probability in predicting the motion vector of the current block can be determined according to the location of the current block in the coding unit. The location of the current block can be expressed using the partition index of the current block. Split indices are assigned to blocks in a coding unit according to a predetermined order. Thus, the motion vector predictor, with a high probability used in predicting the motion vector of the current block, can be determined according to the partition index of the current block. According to FIG. 13A, when the current block is located on the lower side of the coding unit, the motion vector of the current block may be identical or similar to the block motion vector adjacent to the left side of the coding unit, or the block motion vector adjacent to the lower left side of the coding unit. Accordingly, the placement order can, if necessary, be changed so that the candidate predictor of the motion vector, which corresponds to the motion vector of the block adjacent to the left side, or to the motion vector of the block adjacent to the lower left side, is located in front in group C. Since mv_b ' Of the predictor candidates of the motion vector in group C, the predictor is the candidate motion vector, which corresponds to the motion vector of the block adjacent to the left side, the order of mv_b 'and median (mv_a', mv_b ', mv_c') group C, which allows us to reorder group C in the form C = {mv_b ', mv_a', median (mv_a ', mv_b', mv_c '), mv_c', mv_temporal}.

[229] Similarly, when the current block is located on the left side of the coding unit, as shown in FIG. 13B, a motion vector predictor candidate that corresponds to a block motion vector adjacent to the left side of the coding unit and a block motion vector adjacent to the upper side of the coding unit can be used to predict the motion vector of the current block. Since mv_b 'from the motion vector predictor candidates in group C is the predictor of the motion vector candidate that corresponds to the motion vector of the block adjacent to the left side, the order of mv_b' and median (mv_a ', mv_b', mv_c ') changes in group C, which allows reordering the group C in the form C = {mv_b ', mv_a', median (mv_a ', mv_b', mv_c '), mv_c', mv_temporal}.

[230] Referring to FIG. 13C, when the current block is located on the upper side of the coding unit, a motion vector predictor candidate that corresponds to a block motion vector adjacent to the left side of the coding unit and a block motion vector adjacent to the upper side of the coding unit can be used as a predictor of the current motion vector block. Since mv_a 'of the motion vector predictor candidates in group C is the predictor of the motion vector candidate that corresponds to the motion vector of the block adjacent to the upper side, the order of mv_a' and median (mv_a ', mv_b', mv_c ') changes in group C, which allows reordering the group C in the form C = {mv_a ', median (mv_a', mv_b ', mv_c'), mv_b ', mv_c', mv_temporal}.

[231] Referring to FIG. 13D, when the current block is located on the right side of the coding unit, the motion vector predictor candidate that corresponds to the block motion vector adjacent to the upper side of the coding unit can be used to predict the motion vector of the current block. Since mv_c 'of the motion vector predictor candidates in group C is the motion vector predictor candidate that corresponds to the block motion vector adjacent to the upper right side, the order of mv_c' and median (mv_a ', mv_b', mv_c ') changes in group C, which allows reordering the group C in the form C = {mv_c ', mv_a', mv_b ', median (mv_a', mv_b ', mv_c'), mv_temporal}.

[232] The location of the current block in the coding unit as a criterion used to reorder the motion vector predictors is shown as an example. In other words, various criteria can be used to reorder the motion vector predictors. Various criteria used to place a motion vector predictor candidate, which is very likely to be similar to the motion vector of the current block, in front of group C, can be used as criteria for reordering motion vector predictor candidates. A motion vector predictor candidate, which is very likely similar to the motion vector of the current block, is determined based on predetermined information relating to other blocks encoded before the current block, and group C can be reordered according to the definition.

[233] Furthermore, the motion vector predictor candidate, which is highly likely to be similar to the motion vector of the current block, is determined based on other information encoded or decoded with respect to the current block, before encoding the motion vector of the current block, and group C can be reordered according to definition.

[234] In addition, group C can be reordered by eliminating overlapping motion vector predictor candidates. If there are overlapping motion vector predictor candidates in all motion vector predictor candidates, overlapping motion vector predictor candidates are eliminated first, and the need to exclude motion vector predictor candidates can be determined according to equations 1-5.

[235] Returning to FIG. 9, the motion vector encoder 930 encodes motion vector information and motion vector predictor information. The information about the motion vector is the difference of the initial motion vector between the original motion vector of the current block and the initial predictor of the motion vector. Motion vector predictor information is information for indicating a motion vector predictor candidate used to predict a motion vector of a current block among motion vector predictor candidates from which at least one motion vector predictor is excluded. In other words, information for indicating a motion vector predictor of the current block among motion vector predictor candidates that are not excluded on candidate determination unit 920 is encoded as motion vector predictor information.

[236] The difference of the initial motion vector is received from the motion vector estimator 910 and encoded according to a predetermined statistical encoding method. In addition, information is encoded to indicate a motion vector predictor candidate used to predict a motion vector of a current block from motion vector predictor candidates determined by selectively excluding at least one motion vector predictor candidate on a candidate determination unit 920.

[237] When the candidate determination unit 920 determines motion vector predictor candidates by excluding at least one motion vector predictor candidate from all motion vector predictor candidates according to Equations 1-5, information is encoded to indicate a motion vector predictor candidate used for predictions of the motion vector of the current block from certain predictor candidates of the motion vector. The motion vector encoder 930 can index each of the motion vector predictors that are not excluded on the candidate determination unit 920, and statistically encodes the index information as information about the motion vector predictor. Indexing means assigning predefined binary numbers to each of the motion vector predictor candidates, and motion vector predictor information means a binary number assigned to the predictor motion vector candidate used to predict the motion vector of the current block. When one motion vector predictor candidate remains after the candidate determination unit 920 selectively excludes at least one motion vector predictor candidate, the motion vector predictor information may not be encoded separately on the motion vector encoder 930 because the motion vector predictor is candidate, to be used to predict the motion vector of the current block is implicitly determined.

[238] Furthermore, as described in connection with FIG. 13A-13D, the candidate determination unit 920 may index each of the motion vector predictor candidates generated by reordering all the motion vector predictor candidates according to a predetermined criterion and selectively excluding at least one motion vector predictor from all reordered motion vector predictor candidates , and can statistically encode index information.

[239] As a result of the reordering carried out by the candidate determination unit 920, the binary numbers with the smallest number of bits can be assigned to the motion vector predictor candidate, which is highly likely to be used to predict the motion vector of the current block, and thus, the motion vector predictor information can be encoded with high compression ratio.

[240] In FIG. 14 is a block diagram of an apparatus 1400 for decoding a motion vector according to an illustrative embodiment.

[241] The apparatus 1400 for decoding a motion vector shown in FIG. 14 is included in the image decoding apparatus 200 shown in FIG. 2, or the image decoder 500 shown in FIG. 5. According to FIG. 14, the device 1400 for decoding a motion vector includes a motion vector decoder 1410, a candidate determination unit 1420, and a motion vector recovery unit 1430.

[242] A device 1400 for decoding a motion vector decodes the motion vector of the current block when the motion vector of the current block is encoded according to the explicit mode from the above explicit mode and implicit mode.

[243] The motion vector decoder 1410 receives the bitstream for the motion vector of the current block and decodes the received bitstream. The motion vector information included in the bitstream is decoded. The difference of the initial motion vector of the current block is decoded. The difference of the initial motion vector can be decoded according to a predetermined statistical decoding method. The difference of the initial motion vector is the vector difference between the motion vector of the current block and the predictor motion vector candidate used to predict the motion vector of the current block. According to a motion vector decoding method according to an illustrative embodiment, at least one motion vector predictor candidate is excluded from all motion vector predictor candidates according to Equations 1-5, and motion vector predictor candidates are determined. The motion vector predictor predictors are not fixed and are decoded block by block, so that the motion vector predictor candidates can be continuously changed. Accordingly, although the information about the motion vector predictor candidates is the same if the motion vector predictor candidates are not defined, the motion vector predictor candidate used to predict the motion vector of the current block may not be accurately reconstructed.

[244] Thus, the candidate determination unit 1420 determines the motion vector predictor candidates before determining the motion vector predictor candidate used to predict the motion vector of the current block. At least one motion vector predictor candidate from all motion candidate vectors is selectively excluded according to equations 1 to 5, and motion vector predictor candidates are determined. The motion vector predictor predictor, which is not used to predict the motion vector of the current block, is excluded from all candidates determined based on the motion vectors of the blocks included in a previously encoded area adjacent to the current block based on a predetermined evaluation function.

[245] A virtual motion vector is generated based on a predetermined motion vector and from all motion vector predictors and information about the motion vector predictor decoded at the motion vector decoder and the difference of the virtual motion vector, which are the differences between the generated virtual motion vector and another motion vector predictor predictors are calculated for all candidates. The calculated differences of the virtual motion vector are compared with the information about the motion vector decoded on the decoder 1410 of the motion vector, which is information about the difference of the initial motion vector, and a predetermined predictor candidate motion vector is selectively excluded. The result obtained by statistical coding of the differences of the virtual motion vector is compared with the result obtained by statistical coding of the difference of the initial motion vector, which makes it possible to determine the need to exclude the predetermined motion vector predictor candidate. In addition, to improve the accuracy of estimating the result of statistical coding, the result obtained by statistical coding of index information is also evaluated and used to determine the exception. A method for eliminating motion vector predictor candidates is described in connection with Equations 1-5.

[246] In addition, according to another exemplary embodiment, the candidate determination unit 1420 reorders all motion vector predictor candidates according to a predetermined criterion, repeatingly determines an exception for all reordered motion vector predictor candidates according to equations 1-5, and can selectively exclude at least one predictor candidate motion vector. Overlapping motion vector predictors can be excluded from all reordered motion vector predictors, and the determination of exclusion can be repeated in accordance with Equations 1-5.

[247] When a plurality of motion vector predictor candidates from all motion vector predictor candidates remains after the candidate determination unit 1420 excludes at least one motion vector predictor candidate from all motion vector predictor candidates, the motion vector decoder 1410 decodes the information about the motion vector predictor. The motion vector predictor information is decoded according to a predetermined statistical decoding method. The motion vector predictor information is information for indicating a motion vector predictor candidate used to predict the motion vector of the current block from motion vector predictor candidates from which at least one motion vector predictor candidate is excluded. Information for indicating a motion vector predictor candidate used to predict a motion vector of the current block from motion vector predictor candidates that are not excluded at candidate determination unit 1420 is decoded.

[248] When one motion vector predictor candidate remains after the candidate determination unit 1420 excludes at least one motion vector predictor candidate from all motion vector predictor candidates, the remaining motion vector predictor candidate is used to predict the motion vector of the current block , and thus, the motion vector decoder 1410 may not selectively decode the motion vector predictor candidate information.

[249] The motion vector recovery unit 1430 restores the motion vector of the current block based on the motion vector information decoded on the motion vector decoder 1410. The difference of the initial motion vector decoded at the motion vector decoder 1410 and the motion vector predictor candidate used to predict the motion vector of the current block are added together to reconstruct the motion vector of the current block. The motion vector predictor candidate to be used in the motion vector prediction of the current block is determined from the motion vector predictors determined on the candidate determination unit 1420, and the determined motion vector predictor candidate is added to the difference of the initial motion vector. When, as a result of exclusion, at the candidate determination block 1420, there is not one motion vector predictor candidate candidate but a plurality of motion vector predictor candidates, motion vector predictor candidates used to predict the motion vector of the current block can be determined based on information about the motion vector predictor, decoded on the decoder 1410 motion vector.

[250] Since motion vector predictor candidates are determined by candidate determination unit 1420, even if the decoded motion vector predictor information is the same, motion vector predictor candidates used to predict the motion vector of the current block may be motion vectors of neighboring blocks at different positions.

[251] In FIG. 15 is a flowchart illustrating a method for encoding a motion vector according to an illustrative embodiment.

[252] Referring to FIG. 15, in operation 1510, a device for encoding a motion vector estimates the motion vector of the current block and determines a motion vector predictor used to predict the motion vector of the current block from all motion vector predictor candidates. A block identical or similar to the current block is searched for on a plurality of reference images and, as a result of the search, a motion vector is estimated, which is the relative position difference between the current block and the reference block.

[253] Then, the motion vector of the current block is predicted based on the motion vectors of the blocks included in a previously encoded region adjacent to the current block. In other words, the motion vectors of blocks included in a previously encoded region adjacent to the current block are set as motion vector predictors, and a motion vector predictor is determined that is most similar to the motion vector of the estimated current block from motion vector predictor candidates. A vector difference is generated between the motion vector of the current block and the motion vector candidate determined by the predictor, which is the difference of the initial motion vector.

[254] In operation 1520, an image encoding apparatus selectively excludes at least one motion vector predictor candidate from all motion vector predictor candidates. A motion vector predictor candidate that is not used to predict the motion vector of the current block is excluded from all motion vector predictor candidates.

[255] An image encoding apparatus generates a virtual motion vector using a predetermined motion vector predictor candidate from all motion vector predictor candidates and the difference of the initial motion vector generated in operation 1510. The generated virtual motion vector and another motion vector predictor candidate are used to generate the difference of the virtual motion vector. The differences of the virtual motion vector are generated for each of all predictor candidates of the motion vector, and the generated differences of the virtual motion vector are compared with the difference of the initial motion vector, which allows you to selectively exclude a predetermined predictor candidate motion vector.

[256] The process of generating a virtual motion vector and selectively excluding a motion vector predictor candidate in operation 1520 is repeated for all candidates, and thus at least one motion vector predictor candidate can be excluded from all candidates. When the exclusion process is carried out in a repeating manner, the differences of the virtual motion vector are calculated for the motion vector predictor candidates other than the motion vector predictors that are already excluded, and the calculated differences of the virtual motion vector can be compared with the difference of the initial motion vector.

[257] The difference of the virtual motion vector and the difference of the initial motion vector can be estimated and compared with each other based on a predetermined evaluation function, the predetermined evaluation function may be a function that estimates the result of statistical coding. The difference of the virtual motion vector and the difference of the initial motion vector can be compared with each other based on a function that evaluates the result obtained by statistical coding of the difference of the virtual motion vector and the result obtained by statistical coding of the difference of the initial motion vector. In addition, to improve the accuracy of the estimation, the result obtained by encoding the index information can also be evaluated and used in determining the exception. A method for eliminating at least one motion vector predictor candidate from all motion vector predictor candidates is described in connection with Equations 1-5.

[258] Furthermore, as described in connection with FIG. 13A-13D, a motion vector encoding apparatus may reorder all motion vector predictor candidates according to a predetermined criterion and selectively exclude at least one motion vector predictor candidate from all reordered motion vector predictors. In addition, the apparatus for encoding the motion vector can exclude overlapping motion vector predictors from all the reordered motion vector predictors, and repeat the determination of the exception according to Equations 1-5.

[259] In operation 1530, a motion vector encoding apparatus encodes motion vector information and motion vector predictor information. The difference information of the initial motion vector and information for indicating the predictor candidate motion vector used to predict the motion vector of the current block are encoded. The motion vector predictor information may be information to indicate a motion vector predictor candidate used to predict the motion vector of the current block from motion vector predictor candidates that are not excluded in operations 1520 and 1530.

[260] When one motion vector predictor candidate remains after excluding at least one motion vector predictor candidate from all motion vector predictor candidates, the motion vector predictor information may not be encoded.

[261] FIG. 16 is a flowchart illustrating a method of decoding a motion vector according to an illustrative embodiment.

[262] Referring to FIG. 16, a device for decoding a motion vector decodes information about a motion vector of a current block from a received bitstream in operation 1610. Information about a motion vector may be the difference of an initial motion vector between a motion vector of the current block and a motion vector predictor of the current block.

[263] In operation 1620, the motion vector decoding apparatus generates a virtual motion vector based on the motion vector information decoded in operation 1610 and one motion vector predictor candidate from motion vector predictor candidates.

[264] When generating a virtual motion vector, a device for decoding a motion vector excludes at least one motion vector predictor candidate from all motion vector predictor candidates. All motion vector predictor candidates are determined based on the motion vectors of the blocks included in a previously decoded area adjacent to the current block. A device for decoding a motion vector may selectively exclude at least one motion vector predictor candidate from all motion vector predictor candidates. The difference of the virtual motion vector and the difference of the initial motion vector decoded in operation 1610 are estimated based on a predetermined evaluation function to selectively exclude a predetermined motion vector predictor candidate. The method of eliminating the motion vector predictor candidate from all candidates is the same as in operation 1530 and described above with reference to equations 1-5.

[265] The process of generating a virtual motion vector and selectively excluding a motion vector predictor candidate in operation 1620 is repeated for all candidates, and thus at least one motion vector predictor candidate can be excluded from all candidates.

[266] Furthermore, as described in connection with FIG. 13A-13D, a motion vector decoding apparatus may reorder all motion vector predictor candidates according to a predetermined criterion and selectively exclude at least one motion vector predictor candidate from all reordered motion vector predictors. In addition, the motion vector decoding apparatus can exclude overlapping motion vector predictors from all the reordered motion vector predictors, and repeat the determination of the exception according to Equations 1-5.

[267] When a plurality of motion vector predictor candidates remain as a result of the exception, the motion vector predictor information is decoded, and when there is only one motion vector predictor candidate, the motion vector predictor information is not decoded.

[268] In operation 1630, a device for decoding a motion vector determines a motion vector predictor used to predict a motion vector of the current block from motion vector predictor candidates that are not excluded in operation 1620.

[269] The motion vector predictor candidate used to predict the motion vector of the current block from the motion vector predictor candidates can be determined based on the motion vector predictor information of the current block. When, as a result of the exception in operation 1620, there remains one motion vector predictor candidate, one remaining motion vector predictor candidate is determined as the motion vector predictor used to predict the motion vector of the current block.

[270] When the motion vector predictor candidate is determined, the motion vector of the current block is restored by adding up the specific motion vector predictor candidate and the difference of the original motion vector decoded in operation 1610.

[271] According to an illustrative embodiment, when motion vector candidate predictors are used to predictively encode and decode a motion vector, the number of motion vector predictor candidates can be reduced to predictively encode and decode a motion vector. Accordingly, the information necessary to indicate the motion vector predictor used to predict the motion vector of the current block from the motion vector predictor candidates can be encoded with a minimum number of bits, whereby the compression coefficient of the encoding / decoding of the motion vector is increased, which allows to increase the compression ratio when encoding / image decoding.

[272] Although illustrative embodiments are specifically shown and described above, those skilled in the art can propose various changes regarding the form and details without departing from the spirit and scope of the idea of the present invention as defined in the following claims.

[273] Illustrative embodiments may also be implemented as computer-readable codes on a computer-readable recording medium.

[274] For example, an image encoding device, an image decoding device, a motion vector encoder, and a motion vector according to exemplary embodiments may include a bus connected to each element included in the devices of FIG. 1, 2, 4, 5, 9, and 14, and at least one processor coupled to the bus. In addition, devices may include memory connected to at least one processor for executing instructions by combining with a bus for storing instructions, received messages, or generated messages.

[275] A computer-readable recording medium is any data storage device on which data can be stored, which can subsequently be read by a computer system. Examples of computer-readable recording media include read-only memory (ROM), random access memory (RAM), CD-ROMs, magnetic tapes, floppy disks and optical storage devices. A computer-readable recording medium can also be distributed over computer systems connected to a network, which makes distributed storage and execution of computer-readable code possible.

Claims (51)

1. An image decoding device comprising: a decoding unit configured to decode, from the bitstream, information about the difference of the motion vectors for the current block and information about the predictor of the motion vector for the current block; a candidate determination unit configured to generate a group of motion vector predictor candidates and change the said motion vector predictor group based on the values of motion vector predictor candidates in said motion vector predictor group and the number of motion vector predictor candidates in said predictor group - motion vector candidates; and a motion vector recovery unit, configured to determine a motion vector predictor for the current block based on a group of motion vector predictor candidates and said motion vector predictor information, and determine a motion vector for the current block based on said motion vector predictor and said vector difference information movements wherein the modified group of motion vector predictor candidates includes at least one of the first motion vector predictor candidates and a second motion vector predictor candidate, wherein the first motion vector predictor candidates are based on motion vectors of neighboring blocks of the current block, and the second predictor the motion vector candidate is based on the motion vector of the co-located block for the current block, and said co-located block is in the reference image, moreover, neighboring blocks include a first block located lower left of the current block, and include a second block located above the first block. 2. The device according to claim 1, in which the candidate determination unit changes the group of motion vector predictor candidates when there are multiple first motion vector predictor candidates with overlapping values in said group of motion vector predictor candidates by eliminating one of the first candidate predictors motion vectors with overlapping values from the said group of motion vector predictor candidates. 3. The device according to claim 1, in which the candidate determination unit changes the group of motion vector predictor candidates to include a total of two motion vector predictor candidates, and wherein the candidate determination unit determines one of said two motion vector predictor candidates as a motion vector predictor based on information about the motion vector predictor having a 1-bit binary value. 4. The device according to claim 1, in which the second predictor candidate motion vector is scaled based on the time distance between the current image and the reference image. 5. An image encoding method, comprising: determining a motion vector of the current block by performing mutual prediction with respect to the current block; generating a group of motion vector predictor candidates; changing the group of motion vector predictor candidates based on the values of the motion vector predictor candidates in said group of motion vector predictor candidates; encoding motion vector predictor information indicating a motion vector predictor of the current block by determining a motion vector predictor of the current block among motion vector predictor candidates included in a modified group of motion vector predictor candidates; encoding information about the difference of the motion vectors of the current block, which indicates the difference between the motion vector of the current block and the predictor of the motion vector of the current block; and generating a bit stream including information about the difference of the motion vectors of the current block and information about the predictor of the motion vector indicating the predictor of the motion vector of the current block, moreover, the modified group of motion vector predictor candidates includes at least one of the first motion vector predictor candidates and a second motion vector predictor candidate, the first motion vector predictor candidates based on motion vectors of neighboring blocks of the current block, and the second predictor the motion vector candidate is based on a motion vector of a co-located block for the current block, and said co-located block is in the reference image, wherein neighboring blocks include a first block located lower left of the current block, and include a second block located above the first block. 6. A non-transitory computer-readable medium storing a bitstream, comprising: motion vector predictor information indicating a motion vector predictor of the current block; and information about the difference of the motion vectors of the current block, which indicates the difference between the motion vector of the current block and the predictor of the motion vector of the current block, moreover, the motion vector of the current block is determined by performing mutual prediction with respect to the current block, this generates a group of motion vector predictor candidates, said generated group of motion vector predictor candidates changing based on the values of motion vector predictor candidates in a group of motion vector predictor candidates, said motion vector predictor information is information that is encoded by determining a motion vector predictor of the current block among motion vector predictor candidates included in a modified group of motion vector predictor candidates, moreover, the modified group of motion vector predictor candidates includes at least one of the first motion vector predictor candidates and a second motion vector predictor candidate, the first motion vector predictor candidates based on motion vectors of neighboring blocks of the current block, and the second predictor the motion vector candidate is based on a motion vector of a co-located block for the current block, and said co-located block is in the reference image, moreover, neighboring blocks include a first block located lower left of the current block, and include a second block located above the first block. 7. An image encoding device comprising: a motion vector estimator, configured to determine a motion vector of the current block by performing mutual prediction with respect to the current block, a candidate determination unit configured to generate a group of motion vector predictor candidates and change the said motion vector predictor group based on the values of motion vector predictor candidates in said motion vector predictor group; and motion vector encoder, configured to encode motion vector predictor information indicating a motion vector predictor of a current block by determining a motion vector predictor of a current block among motion vector predictor candidates included in a modified group of motion vector predictor candidates, encoding motion vector difference information of the current block, which indicates the difference between the motion vector of the current block and the predictor of the motion vector uschego unit, and generating a bitstream including information about the difference between the motion vectors of the current block, and information on the motion vector predictor, indicating the motion vector predictor of the current block, moreover, the modified group of motion vector predictor candidates includes at least one of the first motion vector predictor candidates and a second motion vector predictor candidate, the first motion vector predictor candidates based on motion vectors of neighboring blocks of the current block, and the second predictor the motion vector candidate is based on a motion vector of a co-located block for the current block, and said co-located block is in the reference image, and moreover, neighboring blocks include a first block located lower left of the current block, and include a second block located above the first block. 8. A non-transitory computer-readable medium storing a bitstream comprising: motion vector predictor information indicating a motion vector predictor of the current block; and information about the difference of the motion vectors of the current block, which indicates the difference between the motion vector of the current block and the predictor of the motion vector of the current block, moreover, the motion vector of the current block is determined by performing mutual prediction with respect to the current block, this generates a group of motion vector predictor candidates, said generated group of motion vector predictor candidates changing based on the values of motion vector predictor candidates in a group of motion vector predictor candidates, said motion vector predictor information is information that is encoded by determining a motion vector predictor of the current block among motion vector predictor candidates included in a modified group of motion vector predictor candidates, moreover, the modified group of motion vector predictor candidates includes at least one of the first motion vector predictor candidates and a second motion vector predictor candidate, the first motion vector predictor candidates based on motion vectors of neighboring blocks of the current block, and the second predictor the motion vector candidate is based on a motion vector of a co-located block for the current block, and said co-located block is in the reference image, and neighboring blocks include a first block located lower left of the current block, a second block located above the first block, and blocks located upper left, upper right of the current block. 9. A method for decoding an image, comprising: decoding, from the bitstream, the motion vector difference information for the current block and the motion vector predictor information for the current block; generating a group of motion vector predictor candidates; changing the group of motion vector predictor candidates based on the values of motion vector predictor candidates in the motion vector predictor group and the number of motion vector predictor candidates in the motion vector predictor group; determining a motion vector predictor for the current block based on a group of motion vector predictor candidates and motion vector predictor information; and determining a motion vector for the current block based on the predictor of the motion vector and information about the difference of the motion vectors, moreover, the modified group of motion vector predictor candidates includes at least one of the first motion vector predictor candidates and a second motion vector predictor candidate, the first motion vector predictor candidates based on motion vectors of neighboring blocks of the current block, and the second predictor the motion vector candidate is based on a motion vector of a co-located block for the current block, and said co-located block is in the reference image, wherein neighboring blocks include a first block located lower left of the current block, and include a second block located above the first block.

Images