KR20230028745A - Method and apparatus for video encoding/decoding using image analysis - Google Patents

Abstract

Disclosed are a method and an apparatus for encoding/decoding a video based on image characteristics according to image analysis. The method for encoding a video using image analysis, includes the following steps of: classifying pixels included in a current block based on a preset band section; analyzing image characteristics of the current block using a classification result for the pixels; and performing intra prediction on the current block in accordance with the analysis result about the image characteristics. Therefore, the present invention is capable of improving encoding efficiency as well as reducing complexity through infra prediction applying a reversal motion to an image having image characteristics different from that of a natural image such as screen content, and in-loop filtering applying an alternative value.

Description

Video encoding/decoding method and apparatus using image analysis {METHOD AND APPARATUS FOR VIDEO ENCODING/DECODING USING IMAGE ANALYSIS}

The present invention relates to video encoding/decoding, and more particularly, to a method and apparatus for encoding/decoding a video based on image characteristics according to image analysis.

Standardization of HEVC (High Efficiency Video Coding), which is known to have about twice or more compression efficiency compared to conventional H.264/AVC, has been recently completed.

HEVC defines a Coding Unit (CU), Prediction Unit (PU), and Transform Unit (TU) with a quadtree structure, and Sample Adaptive Offset (SAO). Offset) and in-loop filters such as a deblocking filter are applied. In addition, compression encoding efficiency is improved by improving existing intra prediction and inter prediction.

In the case of conventional intra prediction, it is difficult to perform accurate prediction according to the characteristics of an image. In particular, when a prediction block is generated using an extrapolation method, generally, as the distance between the prediction pixel and the reference pixel increases, the prediction accuracy decreases.

In addition, when a pattern such as a certain random shape or line repeatedly appears in a screen, accurate prediction cannot be performed by conventional intra prediction either. In order to overcome this problem, prediction is performed using motion search and compensation in the same way as the block matching algorithm used in inter prediction, or a template constructed using neighboring reference pixels of the current block is used. Motion detection and compensation can be performed using Furthermore, the accuracy of prediction can be improved if it is performed using another method having a similar purpose to the above-mentioned methods.

For example, in the case of a block matching algorithm, an optimal encoding cost (based on similarity between pixels, SAD (sum of absolute absolute difference), sum of absolute transformed difference (SATD) or sum of the squared difference (SSD) or other methods may be used, and rate-distortion methods such as RDO (rate distortion optimization) similarity and occurrence between pixels A method of deriving a cost in consideration of bits and the like may be used.

However, when a prediction block is generated using motion search and compensation as a method of intra prediction, prediction accuracy may be lowered in an area having some image characteristics, such as screen content.

In addition, there is a problem in that the efficiency of the SAO technique is low in an area having some image characteristics such as screen content.

An object of the present invention to solve the above problems is to provide a method of adaptively encoding and decoding a video through analysis of the video.

Another object of the present invention to solve the above problem is to provide an apparatus for adaptively encoding and decoding a video through analysis of the video.

To achieve the above object, a video encoding method using image analysis according to an embodiment of the present invention includes the steps of classifying pixels included in a current block based on preset band classification; analyzing image characteristics of a current block by using a classification result for a pixel; and performing intra prediction on the current block according to an analysis result of image characteristics.

Here, in the step of classifying the pixels included in the current block, a classification result for the pixels may be calculated in the form of an index map including information on a band to which the luminance of the pixel belongs and information on a frequency occupied by the corresponding band. there is.

Here, the analyzing of the image characteristics of the current block may include at least one of whether or not an edge exists in the current block, whether the current block is a background area, and whether the current block is a screen content area using a classification result for pixels. one can be analyzed.

In the performing of intra prediction on the current block, the intra prediction is performed on the current block by determining whether to apply an inversion motion to intra prediction based on block matching or template matching based on an analysis result of image characteristics. can do.

In the performing of intra prediction on the current block, when an edge exists in the current block or when the current block is a screen content area, intra prediction may be performed on the current block by applying an inversion motion.

Here, the method may further include performing in-loop filtering on a reconstructed block obtained by reconstructing the current block according to an analysis result of image characteristics.

Here, in the step of performing the in-loop filtering, filtering may be performed using a Sample Adaptive Offset (SAO) or a replacement value.

Here, in the performing of the in-loop filtering, when an edge exists in the current block or when the current block is a screen content area, filtering may be performed by applying a replacement value to a restored block obtained by reconstructing the current block. there is.

To achieve the above object, a video decoding method using image analysis according to an embodiment of the present invention includes determining a decoding method for a current block according to an analysis result of image characteristics of the current block; and performing intra prediction on the current block by determining whether to apply inversion motion based on a decoding method for the current block according to an analysis result of image characteristics.

Here, the analysis result of the image characteristics may be calculated by analyzing at least one of whether an edge exists in the current block, whether the current block is a background area, and whether the current block is a screen content area.

In the performing of intra prediction on the current block, the intra prediction is performed on the current block by determining whether to apply an inversion motion to intra prediction based on block matching or template matching based on an analysis result of image characteristics. can do.

In the performing of intra prediction on the current block, when an edge exists in the current block or when the current block is a screen content area, intra prediction may be performed on the current block by applying an inversion motion.

Here, the method may further include performing in-loop filtering on the current block based on a decoding method for the current block according to an analysis result of image characteristics.

Here, in the step of performing the in-loop filtering, filtering may be performed using a Sample Adaptive Offset (SAO) or a replacement value.

In the performing of the in-loop filtering, when an edge exists in the current block or when the current block is a screen content area, filtering may be performed by applying a replacement value to the current block.

In order to achieve the above object, a video decoding apparatus using image analysis according to another embodiment of the present invention includes an entropy decoding unit that calculates information on a decoding method for the current block according to an analysis result of image characteristics of the current block. and; and an intra predictor for performing intra prediction on the current block by determining whether to apply inversion motion based on a decoding method for the current block according to a result of analyzing the image characteristics.

An apparatus for performing image analysis according to another embodiment of the present invention for achieving the above object includes: a pixel classification unit for classifying pixels included in a current block based on classification of a preset band; an image analysis unit that analyzes image characteristics of the current block using a classification result for pixels; and a mode addition determining unit that determines whether a new mode needs to be added according to an analysis result of image characteristics.

Here, the mode addition determining unit determines whether to perform intra prediction on the current block by applying an inversion motion when an edge exists in the current block or when the current block is a screen content area, and a sample adaptive offset (SAO: Sample Adaptive Offset) or alternative values can be used to determine whether or not to apply filtering.

The video encoding/decoding method using image analysis according to the present invention as described above is performed through intra prediction using inversion motion and in-loop filtering using replacement values for images having image characteristics different from natural images, such as screen content. Encoding efficiency can be improved and complexity can be reduced.

1 is a block diagram illustrating an apparatus for performing image analysis according to an embodiment of the present invention.
2 is an exemplary view for explaining an index map generated through pixel analysis according to an embodiment of the present invention.
3 is an exemplary view for explaining a case in which a part of an image is inverted according to an embodiment of the present invention.
4 is an exemplary diagram for explaining a pixel arrangement when an image is inverted according to another embodiment of the present invention.
5 is an exemplary diagram for explaining motion estimation in the case of having inverted image characteristics according to an embodiment of the present invention.
6 is an exemplary diagram for explaining analysis of image characteristics according to an embodiment of the present invention.
7 is an exemplary diagram for explaining image characteristics according to an embodiment of the present invention based on pixel directionality.
8 is an exemplary diagram for explaining an additional mode setting based on pixel band classification according to an embodiment of the present invention.
9 is an exemplary diagram for explaining a sample adaptive offset technique.
10 is an exemplary diagram for explaining in-loop filtering according to an embodiment of the present invention.
11 is an exemplary diagram for explaining parameters used for in-loop filtering according to an embodiment of the present invention.
12 is an exemplary diagram for explaining in-loop filtering according to another embodiment of the present invention.
13 is an exemplary diagram for explaining in-loop filtering according to another embodiment of the present invention.
14 is a block diagram for explaining a video encoding apparatus using image analysis according to an embodiment of the present invention.
15 is a block diagram for explaining a video decoding apparatus using image analysis according to an embodiment of the present invention.

Since the present invention can make various changes and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. Like reference numerals have been used for like elements throughout the description of each figure.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present invention. The terms and/or include any combination of a plurality of related recited items or any of a plurality of related recited items.

It is understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, but other elements may exist in the middle. It should be. On the other hand, when an element is referred to as “directly connected” or “directly connected” to another element, it should be understood that no other element exists in the middle.

Terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, the terms "include" or "have" are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in the present application, they should not be interpreted in an ideal or excessively formal meaning. don't

The video encoding apparatus (Video Encoding Apparatus) and the video decoding apparatus (Video Decoding Apparatus), which will be described later, include a personal computer (PC: Personal Computer), a notebook computer, a personal digital assistant (PDA), and a portable multimedia player (PMP). : Portable Multimedia Player), PlayStation Portable (PSP: PlayStation Portable), wireless communication terminal (Wireless Communication Terminal), smart phone (Smart Phone), it may be a server terminal such as a TV application server and service server, and various devices such as A user terminal or a communication device such as a communication modem for communicating with a wired/wireless communication network, memory for storing various programs and data for encoding or decoding images, inter- or intra-screen prediction for encoding or decoding, and execution of programs It may refer to various devices including a microprocessor and the like for calculation and control.

In addition, the image encoded as a bitstream by the image encoding device is transmitted through a wired or wireless communication network such as the Internet, a local area wireless communication network, a wireless LAN network, a WiBro network, a mobile communication network, etc. in real time or non-real time, or through a cable or universal serial bus (USB: Universal). It can be transmitted to the video decoding device through various communication interfaces such as a serial bus), decoded in the video decoding device, and restored and reproduced as an image.

In general, a moving picture may be composed of a series of pictures, and each picture may be divided into predetermined areas such as frames or blocks. When an area of an image is divided into blocks, the divided blocks may be largely classified into intra blocks and inter blocks according to encoding methods. An intra block refers to a block that is coded using the intra prediction coding method. This is a method of generating a prediction block by predicting pixels of the current block and encoding a difference between the pixels of the current block and the pixels of the current block. Inter-block refers to a block coded using inter-prediction coding. Inter-prediction coding refers to one or more past or future pictures to predict a current block in a current picture to generate a predicted block and generate a current block. It is a method of encoding the difference value between and. Here, a frame referred to for encoding or decoding a current picture is referred to as a reference frame. In addition, those of ordinary skill in the art to which this embodiment pertains that the term "picture" described below may be replaced with other terms having equivalent meanings such as image and frame. When you grow up, you will understand. Also, in the present invention, a reference picture means a reconstructed picture, which can be understood by those skilled in the art to which this embodiment belongs.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram for explaining an apparatus for performing image analysis according to an embodiment of the present invention, and FIG. 2 is an exemplary diagram for explaining an index map generated through pixel analysis according to an embodiment of the present invention.

Referring to FIG. 1 , an apparatus 10 for performing image analysis according to an embodiment of the present invention includes an image characteristic candidate selection unit 11, a pixel classification unit 12, an image analysis unit 13, and a mode addition determination unit. (14), an additional mode execution unit 15, an additional mode identification flag and related parameter encoding unit 16.

In addition, the device 10 for performing image analysis according to an embodiment of the present invention may be implemented by being included in a video encoding device.

The video feature candidate selection unit 11 may select one of the video feature candidate groups for a video belonging to a coding unit having a predetermined size, and select whether to set an additional mode in consideration of the video feature candidate group.

Here, the image characteristic candidate group can be set in various ways, for example, whether or not an area with an edge exists, whether or not an area belongs to the background, whether or not an area where an object exists, artificial images (computer graphics, computer capture screen) etc.), whether it is a natural image, and whether it is an area belonging to a character or figure.

If an image feature candidate to be considered among several image feature candidates is 1. whether or not a region belonging to the background and 2. whether or not an artificial image, 1, 2 through processes such as the pixel classification unit 12 and the image analysis unit 13 is true (that is, it belongs to the background and belongs to the artificial image), the process of applying a new in-loop filtering that can be well applied to the artificial image belonging to the background to the existing filter (existing mode) is set as an additional mode. Among them, an optimal filter can be applied.

In addition, if the image feature candidates to be considered among the image feature candidates are 1. artificial image or not and 2. text area or not, this is also true through the subsequent process, if 1 and 2 are true (ie, artificial image and text area), the existing By adding a new intra-prediction method that can be well applied to artificial images including text areas, the optimal intra-prediction method can be applied.

Of course, various characteristics may be included in addition to the image feature candidate group described above, and the range of additional modes considered according to the image feature candidates can be applied to a prediction unit, a transform unit, a quantization unit, an in-loop filter unit, etc. to be described later.

In addition, information on an additional mode (or a video feature candidate selection flag) considered in association with a video feature candidate may be transmitted in units of sequences, pictures, slices, and the like.

The pixel classification unit 12 performs preparation work for analyzing the image characteristics selected by the image characteristic candidate selection unit 11 . The pixel classification unit 11 may classify pixels according to various criteria, such as luminance, color difference, location, gradient, correlation, or frequency domain coefficient of pixels belonging to a coding unit of a certain size.

For example, the pixel classification unit 12 may classify pixels according to their luminance values, or may classify pixels according to conversion coefficients in the frequency domain. It is also possible to classify pixels according to displacements with neighboring pixels by constructing (eg, expressed in a vector format such as (Y, Cb, Cr) or (R, G, B)).

These criteria may be fixed or adaptively applied according to the resolution of the image, quantization parameter, coding unit and characteristics. That is, pixels included in the current block may be classified based on a preset band division.

For example, in the case of the luminance of a pixel, the number of bands can be divided by 2n-k (k is a constant greater than 0 and smaller than n) according to the bit depth (n) of the pixel.

In addition, preset or transmitted information, quantization parameters, size of coding units, parameters related to image resolution and encoding (block division information, prediction mode information, information on transformation unit size, coded block pattern, etc.), etc. The luminance of the pixel may be distinguished for each band set by . Here, the width of the band may have a uniform interval or may have an unequal interval depending on conditions such as image characteristics.

Also, the number of bands may have an integer of 2 or more. For example, when it is assumed that pixels are classified according to luminance values in m bands (assumed to have equal intervals here), a frequency count can be calculated according to which band each pixel belongs to.

Referring to FIG. 2, an index map generated through pixel classification will be described.

For example, if one or more pixels belong to a certain band, NZ_i of the corresponding band may be classified as true, and otherwise classified as false. In addition, it can be classified by a unit (Pi) such as a percentage of the pixel frequency occupied by the corresponding band.

Also, pixels may be classified according to two or more criteria. After classifying NZ_i of each band according to the luminance value of a pixel, a new index map may be constructed by assigning a new index to the corresponding band considering only the band having a true value.

If NZ_i is true for bands 0 and m and false for the other bands, since only two candidates exist, 1 bit can be allocated to assign bands 0 and 1 respectively. That is, a new index map may be configured by assigning 0 to the index of band 0 and 1 to band m.

A pixel classification result may be stored in units of a table-type index map generated through such classification.

In addition, classification based on the characteristics of one pixel, such as classification based on criteria such as luminance, color difference, position, etc. of a pixel, or by grouping them, and classification according to the corresponding characteristics, such as gradient, correlation, etc., between neighboring pixels It is also possible to classify according to correlation.

For example, there are various edge detection algorithms such as Sobel, Robert, and Canny that are processed based on neighboring pixels. In this case, as for neighboring pixels, pixels belonging to an M-sized block centered on the current pixel may be used for edge detection.

The image analysis unit 13 serves to analyze the characteristics of the image selected by the image characteristic candidate selection unit 11 . The characteristics of the image may be variously performed according to the purpose of analyzing the image characteristics selected by the image characteristic candidate selection unit 11 .

The image analysis unit 13 determines whether an edge exists or not, whether an area belongs to the background or not, whether an object exists or not, whether it is an artificial image (computer graphics, computer captured screen, etc.) or a natural image, and letters or figures. It is possible to analyze the image, such as whether it belongs to the region or not.

For example, in order to determine whether an area has an edge, its characteristics may be identified based on information transmitted through the pixel classification unit 12 . The image analyzer 13 may analyze information obtained through the edge detection algorithm of the pixel classifier 12 .

In detail, the image analyzer 13 may calculate an analysis result that the corresponding region has an edge when N or more pixels are determined to be edges. Otherwise, the analysis result may be calculated as not having an edge in the corresponding area.

In addition, when N or more pixels are determined to be edges and have continuity, an analysis result may be calculated as indicating that the corresponding area has an edge. An example of continuity can be understood with reference to the description of FIG. 6 to be described later.

The image analyzer 13 classifies the band according to the luminance component of the pixel from the pixel classifier 12, and then analyzes whether the corresponding region belongs to the background or not by utilizing information on NZ_i.

For example, analysis may be performed with the criterion of background only when the sum of NZ_i has a value of N or less. If the sum of NZ_i has a value of N or less, a result that the corresponding region is a background may be calculated, and otherwise, a result of not being a background may be calculated.

Furthermore, the image analyzer 13 may perform image analysis based on two or more pieces of information. That is, after analyzing different image characteristics from one piece of information, a new analysis result can be obtained through each image analysis result. For example, when the pixel classification unit 12 classifies the area into bands according to the luminance component of a pixel in order to analyze whether the corresponding area includes a text portion, analysis is performed using information on NZ_i and information on Pi. can do. In this case, whether or not it is a character part can be defined as a character part only when NZ_i is N or less (here, set to 2) and Pi of the band having the lowest frequency is Th or less. Therefore, by combining the analysis results according to each condition, if both conditions are true, the analysis result is that the corresponding area is an area containing letters, and if either condition is false, the analysis result is that the corresponding area is not a letter area. can be calculated.

The mode addition determination unit 14 determines whether a new mode needs to be added based on one or more analysis results transmitted from the image analysis unit 13, and determines whether to support the additional mode based on the result.

For example, when determining whether or not to generate additional information related to an inversion motion, which will be described later, it is first determined whether the image region is likely to have an inversion motion, and then, if it is determined that there is a possibility of an inversion motion, a mode related to the inversion motion is selected. Add, and if it is determined that it is not, the same algorithm as before can be performed.

In order to confirm that there is a high possibility that there is a reversal motion, the mode addition determining unit 14 determines whether the region has an edge from the image analysis unit 13 or a figure (here, it is assumed that the region specifically includes the vertex portion of a rectangle). An analysis result such as whether a region belongs to may be delivered.

When the additional mode execution unit 15 confirms that there is a high possibility of inversion motion from the mode addition determination unit 14 and decides to add a new prediction mode, the new prediction mode considering the inversion motion is applied to perform encoding can do.

The additional mode identification flag and related parameter encoding unit 16 may transmit an additional mode identification flag generated by adding a new prediction mode in the additional mode executing unit 15 and parameters related thereto.

For example, a new prediction mode such as a block matching algorithm to which inversion motion is applied has been added, and when this algorithm is optimal, inversion motion-related information and motion vector-related information may be transmitted together.

In the above example, only a description related to the inversion motion is presented, but other cases will be described through the following embodiments.

3 is an exemplary diagram for explaining a case in which a portion of an image is inverted according to an embodiment of the present invention, and FIG. FIG. 5 is an exemplary diagram for explaining motion estimation in the case of having an inverted image characteristic according to an embodiment of the present invention.

Referring to FIG. 3 , unlike a general natural video, a screen content image may have inverted characteristics.

In detail, the image shown in FIG. 3 may have characteristics such as left-right inversion and up-down inversion.

For example, in FIG. 3 , a1<->a2,b may have a horizontally inverted characteristic, and a1<->a3,c may have a vertically inverted characteristic. Furthermore, a1<->a4, d may have characteristics of being inverted together up and down, left and right.

However, it is difficult to select a region having such a reversal characteristic as a prediction block through conventional motion search and compensation, even though it has high utility value as a prediction block. If motion search and compensation are performed through transformations such as horizontal inversion and vertical inversion, encoding efficiency can be increased because more accurate motion search and compensation are possible.

The arrangement of pixels will be described with reference to FIG. 4 .

4(a) shows a general arrangement of pixels. If it is assumed that motion search is performed by performing vertical inversion, prediction can be performed by performing pixel rearrangement in the scan order as shown in FIG. 4 (b).

4(c) and 4(d) show cases of left-right inversion and up-down-left-right inversion, respectively. When motion prediction is performed through the configuration of the target block through such pixel rearrangement (y-a y-b, y-c, y-d) and a candidate having the optimal coding efficiency is selected, in the case of the rest except for FIG. 4 (a), the motion vector Compensation for the current block may be performed through reverse pixel rearrangement (y-b, y-c, y-d y-a) of the corresponding block.

In the case of a block in which an inversion has occurred, the decoder may perform compensation for the current block through reverse pixel rearrangement of the block corresponding to the motion vector. Figure 4 is only a few examples of case inversion, the present invention also includes variations for similar purposes.

In addition, the same prediction can be performed by information about application of inversion by information pre-determined on both sides so that the encoder and decoder perform the same operation, or by transmitting inversion information in units of pictures, slices, blocks, etc. For example, an inversion candidate to which inversion is applicable may be previously set according to the block size, or inversion may not be applied according to the block size.

For example, when a method of generating a prediction block by using motion search and compensation as a method of intra prediction and generating a prediction block through motion search and compensation by inversion is used, in an area having some characteristics of an image, Encoding efficiency can be increased.

When motion search and compensation are performed by applying inversion, a prediction vector to be used for motion vector prediction can be set through various methods for the purpose of reducing data when encoding a motion vector.

Motion search and compensation will be described with reference to FIG. 5 .

A block or motion vectors of blocks most recently encoded through motion search and compensation may be selected as a candidate for a prediction vector (A in FIG. 5).

In addition, the relative position of the fixed coordinates (for example, the upper left coordinate of the left block or the upper left coordinate of the upper block) with the current block (if the horizontal and vertical dimensions of the block are v and w, respectively (-v, 0), (0, -w)) can be selected as a candidate group for the prediction vector (B in FIG. 5).

In addition, a motion vector of a block or blocks coded through motion search and compensation to which inversion is most recently applied may be selected as a candidate for a predictive vector, and a block or blocks coded through motion search and compensation to which inversion is applied most recently may be selected. A position relative to the current block may be selected as a candidate for a prediction vector (C in FIG. 5).

In addition, the upper left coordinates of a picture, slice, block, etc. may be selected as a predictive vector candidate. (D in Fig. 5)

Each of the prediction candidates described above can be configured in various combinations, and the priority order for constructing the candidate group of predictive vectors may be determined according to preset conditions of an encoder and a decoding period, and accordingly in units of sequences, pictures, slices, etc. information can be set.

For example, in FIG. 5, when A, B, C, etc. are possible as candidate groups of prediction vectors, it is possible to use only the candidate in C, use the candidates of C and B, or use the candidates of A and B. can

When performing motion search and compensation in an area where encoding or decoding prior to the current block is completed, if it is determined that there is a high possibility that a prediction block having a motion by inversion exists according to a preset rule, inversion information is sent, and otherwise, it is omitted. can do.

For example, if it is 0, it may mean that the existing motion search and compensation are performed without inversion, and there is no need to transmit additional information accordingly. In addition, if it is 1, it may mean that a motion with inversion is searched for and compensated for, and additional information about the resulting inversion is transmitted. However, when it is determined that reversal information does not exist, even information such as 0 and 1 regarding whether reversal information is used may not be transmitted, and through this, encoding performance may be improved.

6 is an exemplary diagram for explaining analysis of image characteristics according to an embodiment of the present invention.

Referring to FIG. 6 , the image analyzer 13 may receive NZ_i and an index map from the pixel classifier 12 and analyze whether the image is a solid segmentation region or a complex region. FIG. 6 (a) may indicate a solid segmented region, and FIG. 6 (b) may indicate a complex region.

There may be two conditions for determining a solid partitioned area.

The first condition is that NZ_i must be less than k (here, assumed to be 5), and the second condition is that the continuity condition of pixels in each area must be satisfied, and the pixels in each area must not be divided into two or more areas. It can be defined as a condition that must be satisfied.

Here, the definition of continuity is as follows.

For example, pixels divided into each band, as shown in area c of FIG. 6 (1), may have the same index map. Here, the definition of continuity is based on the relationship with neighboring pixels in a<x<a, -b<y<b (where x, y are the positions of the current pixel on a two-dimensional plane) centered on the current pixel can be defined as determined.

If it is assumed that there is at least one pixel with the same index map as the current pixel among the neighboring pixels in a<x<a and -b<y<b, there is continuity, and if not, there is no continuity. can be calculated.

Therefore, in the case of region c of FIG. 6(1) (here, it is assumed that the indices of a, b, and c are '0', '1', and '2'), all pixels in the region having an index of 0 are in continuity. , and since it is possible to calculate the analysis result that all pixels with an index of '1' also have continuity, and that pixels in an area with an index of '2' also have continuity, it is possible to calculate that all pixels in the corresponding area have continuity. An analysis result can be obtained, and since there is no area divided into two or more areas, an analysis result that the corresponding area is composed of solid division can be obtained.

In comparison, in the case of region c of FIG. 6 (2) (here, assuming that the indices of a, b, c, and d are '0', '1', '2', and '3'), 0, '1 It is possible to calculate an analysis result that all pixels in the region having indices of ', '2', and '3' have continuity.

However, the above condition is not satisfied because the remaining regions except for the region having an index of '3' are divided into two or more regions. Therefore, in this case, the first and second conditions are satisfied, but the third condition is not satisfied, so that analysis that the area is made of solid division cannot be obtained.

As such, conditions for image characteristics may be defined by the user. Analysis results can be obtained by checking all three conditions, and it is also possible to set one condition to be checked first and proceed to the step of checking the next condition only when it is satisfied.

7 is an exemplary diagram for explaining image characteristics according to an embodiment of the present invention based on pixel directionality.

Referring to FIG. 7 , a case in which an analysis is performed as to whether an area belongs to a figure (here, it is assumed that the area includes a vertex of a rectangle, that is, a bent portion) will be described.

In order to analyze whether the area is a figure, NZ_i of a band according to pixel classification, index map, and edge information may be received from the pixel classification unit 12 .

For example, the condition that NZ_i must be less than D, the condition that pixels in each region must satisfy the continuity condition, the condition that more than N pixels turn out to be edges, the condition that more than L pixels have continuity in the edge index map It can be defined as producing the analysis result that there is a possibility that it is a figure area if the condition, the condition that there must be K or more bent points, etc. is satisfied.

After assuming that all other conditions are satisfied, let's check the condition for the bent point.

In the index map for edges (assuming there are only two 0 and 1 here, assuming that 1 is the edge part), whether or not there is a break point is checked for pixels classified as edges. At this time, A (-a <x<a, -b<y<b) You can check the directionality with neighboring pixels.

If the directionality is confirmed within -2<x<2 and -2<y<2, several candidate groups such as 180 degrees, 135 degrees, 90 degrees, and 45 degrees can be placed for the direction as shown in FIG. In addition, if the directionality is confirmed within 3<x<3 and -3<y<3, candidates for more diverse directions can be added.

In addition, it is possible to check whether a bent point exists by checking the relationship with the directionality of pixels classified as neighboring edges. If the difference in directionality between the pixel classified as the previously located edge and the current pixel (here, when the difference is calculated by an angle) is less than d, a broken point has occurred. Otherwise, it can be said that a broken point does not occur. For example, when d is set to 30 degrees, it is possible to calculate an analysis result that no break point occurs because all cases in FIG. 7 are greater than 30 degrees. When d is set to 135 degrees, Fig. 7 (a) results in that no break point occurs because the condition is not satisfied, whereas in Fig. 7 (b), (c), (d), the break point is obtained because the condition is met. get the result that it happened.

If you want to perform image analysis on it to check whether it is the same area as the vertex of the rectangle, to set conditions to characterize it (for example, or if you want to know if it is an area that has the same shape) It can be configured so that analysis results such as true only when a break point occurs only once in , and false otherwise.

In the case of a rectangle, since the part for the broken point will have an angle of 90 degrees, it is possible to analyze this by changing the condition in the part for checking the above broken point (a broken point occurs only when d is 90). there is.

An example of adding a new mode from the above-described embodiment will be described in more detail as follows.

Referring back to FIG. 1 , the image feature candidate selector 11 tries to determine whether a new addition mode is to be applied to intra prediction in consideration of 1. whether or not it belongs to an artificial image and 2. whether it is a solid segmented region. To this end, the pixel classification unit 12 may receive NZ_i and an index map of a band for pixel classification.

The image analyzer 13 can analyze whether the condition that NZ_i must be smaller than D, the continuity condition of pixels in each area, and the condition that pixels in each area must not be divided into two or more areas is satisfied. .

If the image analyzer 13 assumes that all of the above conditions are satisfied, an analysis result indicating that the segmented region belongs to the artificial image can be calculated.

Through this, the mode addition determination unit 14 determines to add an intra prediction mode in which performance is exhibited in a region having a solid segmentation belonging to an artificial image, in addition to the existing intra prediction mode.

8 is an exemplary diagram for explaining an additional mode setting based on pixel band classification according to an embodiment of the present invention.

Referring to FIG. 8, when pixels are classified for each band as shown in FIG. 8 (a) and a new index is assigned to a band for which NZ_i is true, the corresponding block is solidly segmented belonging to an artificial image through the image analysis unit 13. Let's say we get an analysis of an area with .

The additional mode executing unit 15 executes a new intra prediction mode in addition to the existing intra prediction mode. For example, a method of finding a similar block in a previously coded region through a block matching algorithm may be added.

In addition, a method of copying neighboring pixels in units of pixels or pixels may be added. If, in FIG. 8 (a), when the coordinates of the top left of the block are (0,0), the pixel at the position of (1,0) is the pixel (0,0) before the scan (assuming raster scan here). Since the index (0) is the same as ), mode information called CL (copy left) may be transmitted. When a value is copied from a pixel above the current pixel, such as the location of (7, 1), this may be expressed by transmitting mode information called CT (copy top).

In addition, it is also possible to encode the number of repetitions in succession. In order to encode the first row in FIG. 8 (a), the index (0) that occurs first may be coded, and the number of repeated indices appearing consecutively following it may be coded.

Here, the number of consecutive indexes may be limited per row (index (0) + consecutive number (7)) or may not be limited per row (index (0) + consecutive number (10)).

Also, several pixels may be copied. For example, since the pixels from (0,1) to (2,1) are the same as the pixels from (0,0) to (2,0), (CT(copy top), 3(total number to copy) ) can also be coded in the same way. Copying may be performed from additional directions other than CL and CT.

Also, since the pixels from (0,2) to (4,2) are the same as the pixels from (1,1) to (5,1), (CTR (copy top right), 5 (total number to copy) and Encoding can be performed in the same way.

Also, copying may be performed in row units (or column units depending on the scanning direction). For example, the 8th row (from (0,7) to (7,7)) is the same as the 7th row (from (0,6) to (7,6)), so CTL (copy top line) and This may be expressed by transmitting the same mode information.

In addition, since the pixels from (0,1) to (3,1) in FIG. 8 (b) are the same as the result of horizontally inverting the pixels from (4,1) to (7,1), CM (copy mirror) This can be expressed by transmitting mode information such as For this purpose, since the left and right lengths must be the same, this mode information can be encoded only at the starting point in the middle of the row.

Another example of using inversion may be the case in FIG. 8 (c). Unlike FIG. 8 (b), in FIG. 8 (c), many upside down inverted columns (columns 2 to 6) can be found. In the case of the basic scan direction (assuming raster scan), as described above, since encoding using inversion is impossible, pixel rearrangement as shown in FIG. 4 can be performed.

That is, when rearranging pixels such as rotation by 90 degrees as shown in FIG. 8(d), encoding may be performed using the aforementioned CM mode.

Although the foregoing provides various examples of the intra-prediction method, it is not limited to the above-described embodiment and may include various modifications having similar purposes.

In addition, through the mode addition determination unit 14, an analysis of whether 1. belongs to an artificial image or not and 2. whether it is a solid segmentation area is received from the image analysis unit 13, and whether or not a new intra-prediction mode is added. , it can also be considered whether to add other modes.

If the analysis result of the solid segmentation belonging to the artificial image is obtained from the image analysis unit 13, a new mode for transform or quantization as well as the corresponding intra-prediction method may be added. there is.

After performing the above-described new intra-prediction, performing the existing transformation and quantization, and the reverse process, the pixels in each band are mapped to one pixel representing the band without restoring the pixels in each band to actual pixel values. Restoration can be performed by performing the process of In this case, the transform or quantization process may be omitted. That is, not only the new intra-prediction method is considered, but a new mode of transformation and quantization (e.g., no transformation, no quantization, no both, or adaptive transformation). Whether or not to do so, etc.) can also be determined by the mode addition determining unit 14.

M1: New In-Screen Prediction + Transformation + Quantization

M2: new intra-prediction + quantization

M3: New In-Screen Prediction

Addition of modes in various combinations as described above may be considered. In the case of M1, when a residual component is obtained through new prediction, coefficients obtained by performing transformation and quantization processes may be transmitted.

In the case of M2, quantized coefficients of actual pixel values may be transmitted after prediction is performed for indices (2, 3) having a small frequency in FIG. 8 (a), and prediction is performed only for index (3) having the lowest frequency. Then, quantized coefficients of actual pixel values may be transmitted.

In the case of M3, encoding may be performed by transmitting information on a representative value of each band without separately sending a coefficient for a residual component by mapping pixels in each band to a representative value of each band after performing prediction.

In the above description, through the mode addition determination unit 14, 1. analysis of whether or not it belongs to an artificial image and 2. whether it is a solid segmentation area is performed through the image analysis unit 13 for the encoding as in the above example. Although explained, this is only an example and is not limited to the above case.

Referring to FIG. 7 , a case in which an analysis is performed as to whether an area belongs to a figure (here, it is assumed that the area includes a vertex of a rectangle, that is, a bent portion) will be described.

In order to analyze whether the area is a figure, NZ_i of a band according to pixel classification, index map, and edge information may be received from the pixel classification unit 12 .

For example, the condition that NZ_i must be less than D, the condition that pixels in each region must satisfy the continuity condition, the condition that more than N pixels turn out to be edges, the condition that more than L pixels have continuity in the edge index map It can be defined as producing the analysis result that there is a possibility that it is a figure area if the condition, the condition that there must be K or more bent points, etc. is satisfied.

As in the many examples described above, one or more image characteristics can be defined through various methods, and the definition of each image characteristic may follow a rule set in the encoder and decoder, or the encoder generates information about the definition of the characteristic. and transmit it to the decoder. That is, information may be divided and transmitted in detail for each image characteristic, and transmission units may be transmitted in units such as sequences, pictures, slices, blocks, and the like.

9 is an exemplary diagram for explaining a sample adaptive offset technique.

Referring to FIG. 9, in the case of a band offset, after classifying into bands having similar luminance values within a corresponding block, offsets are transmitted for four consecutive bands among them (transmission offset). A pixel value of a reconstructed image may be corrected by adding the transmitted offset value to pixels within a band in which the offset is transmitted.

10 is an exemplary diagram for explaining in-loop filtering according to an embodiment of the present invention, FIG. 11 is an exemplary diagram for explaining parameters used for in-loop filtering according to an embodiment of the present invention, and FIG. It is an exemplary diagram for explaining in-loop filtering according to another embodiment of the present invention, and FIG. 13 is an exemplary diagram for explaining in-loop filtering according to another embodiment of the present invention.

Referring to FIG. 10 , unlike the conventional case of adding a pixel-by-pixel offset to pixels within a specific band of luminance (or color difference) values in order to reduce errors with original image data, pixels within a specific band to pixels within a specific band can be replaced with a value within or near a specific band.

For example, in the case of some images (computer capture screens), unlike natural images, they are often classified as values near some luminance components. Therefore, as shown in FIG. 10 , encoding performance can be improved by substituting values in a corresponding band with specific bands instead of pixel-by-pixel correction.

Specifically, unlike the conventional transmission of information on the starting position of consecutive bands to the decoder, information on the starting position of continuous bands performing filtering is transmitted and the starting positions of individual bands each performing filtering are transmitted. It is optional among sending information about , and finding out information on the start position of consecutive bands performing filtering or the start position of individual bands each performing filtering in the encoder and decoder through analysis of reconstructed pixels. can be set to

The encoder and decoder may promise to use either sending the starting positions of continuous bands, sending the starting positions of individual bands, or automatically finding the starting positions of the bands through analysis of restored pixels. , it is also possible to transmit selection information in units of pictures, slices, or blocks. For example, if the bit depth is 8 and the width of the band is 8, the total number of bands is 32.

The information on the starting position of the band may be an absolute position in the entire band or a relative position according to various conditions. For example, it may be a relative position from a position determined according to an image characteristic. In this case, when an image characteristic is a case where pixels are concentrated in a certain band, the minimum or maximum value in the block before applying the in-loop filter is determined. The starting position of the band including the pixels may be the reference, the starting position of the band having the average value of the luminance component of the pixels of the corresponding block may be the reference, and the starting position of the band with the highest frequency may be the reference. It could be.

In addition, the band width may also have various values rather than a fixed value. The band may be divided in proportion to the bit depth (eg, when the bit depth is 8, the width is 8), or the band may be divided into bands having an adaptive width according to a quantization parameter.

Furthermore, when the characteristic of the image is that pixels are concentrated in a certain band, the bandwidth (for example, when the existing is 8, the size of less than 8 for more precise filtering (when pixel values are corrected through offset) It is also possible to improve the performance of filtering (here corrected through pixel value offset) by reducing the width with .

In the case of sending the starting position of an individual band, the position of each band may be sent as an absolute position in the entire band, or position information may be sent through a relative difference value of each band. For example, the position information of each band may be sent as an absolute position of the entire band, the start position of the first band is sent as an absolute position of the entire band, and the start position of the second band is based on the start position of the first band. As a result, the relative difference may be transmitted as location information.

Also, the location of each band may be a relative location according to various conditions. For example, a position relative to a position determined according to image characteristics may be exemplified. Here, the image characteristic may be the starting position of a band including a pixel having a minimum or maximum value in a corresponding block before applying an in-loop filter when pixels are concentrated in a specific band, and the pixels of the corresponding block It may be a starting position of a band having an average value of luminance components or a starting position of a band having the highest frequency.

In addition, it is possible to select from several candidate groups for the starting position. It is possible to determine starting location information around which point to generate start position information by comparing encoding costs, such as candidate group 1's initial starting point in the entire band and candidate group 2's starting point of the band with the highest frequency of the corresponding block. The candidate group with the highest selectability among multiple candidates is set as the prediction candidate, and if the actual start position and the candidate judged to have the highest selectability match, it becomes true, so 1 bit can be transmitted. Information on the remaining candidates may be generated and transmitted.

Filtering-related parameters (offset or replacement value) are information for minimizing the error between the reconstructed image and the original image of the corresponding band, and are transmitted for each band. It can be transmitted in one of the methods. For example, they are mixed and used in units such as blocks, or only one method is applied.

If the in-loop filter (here, SAO) configures various selection trees, filtering may be performed. Filtering may not be performed, correction may be performed through edge offset, correction may be performed through band offset, or correction may be performed through replacement values. In addition, it may be performed in a new method with a similar purpose other than this.

This setting may be configured by the encoder and decoder under an agreement, or may be configured by transmitting information on which selection tree to configure. That is, this information can be transmitted in units of pictures, slices, or blocks.

For example, it is possible to configure filtering not performed + edge offset, filtering not performed + band offset + edge offset, filtering not performed + edge offset + replacement value, and the like.

In addition, the occurrence probabilities may be different according to encoding parameters of neighboring blocks. For example, if the in-loop filter is applied to all neighboring blocks with edge offsets, the current block has a high probability of applying correction through edge offsets, so the probability of edge offsets occurring is set higher than other methods, so that entropy encoding through this is more efficient. Encoding performance may be improved by generating fewer bits.

As another example, if an in-loop filter is applied to the left block as a band offset and the upper block as a replacement value, the frequency of occurrence is defined in advance to set the probability of occurrence of each occurrence according to these combinations so that entropy encoding can be performed through this. there is.

In addition, this may be set through other encoding parameters. If the left block applies the band offset and the upper block applies the in-loop filter as a replacement value, other occurrence probabilities may be set by checking the prediction mode (herein, limited to the intra prediction mode) of the corresponding block.

If the prediction mode of the corresponding block coincides with one of the neighboring blocks, the occurrence probability of the in-loop filter of the neighboring block can be increased. For example, if the prediction mode of the left block and the prediction mode of the current block are the same, the occurrence probability of the band offset is increased, and if the prediction mode of the upper block and the prediction mode of the current block are the same, the occurrence of the replacement value Increase the probability, and if it is not both, you can set another probability of occurrence.

In addition, different probability information can be set according to the type of picture or slice. For example, probability information may be set differently according to an inter picture or an intra picture.

Furthermore, by copying and utilizing all or part of parameter information of an in-loop filter of a neighboring block, the amount of in-loop filter parameter information of a corresponding block can be reduced to increase coding efficiency.

Parameters used for in-loop filtering according to an embodiment of the present invention will be described with reference to FIG. 11.

For example, assume that a neighboring block uses a band offset and has parameter information as shown in FIG. 11 (a).

If the current block applies the in-loop filter through the band offset, the parameter information of the neighboring block is fully copied through the 1-bit flag indicating full copy of the in-loop filter parameter information from the neighboring block when '1'. In the case of 0', it is possible to obtain parameter information of the current block from a neighboring block by selecting one of non-copying.

In addition, parameter information of the current block may be obtained from the neighboring block by partially copying the parameter information of the neighboring block. In this case, through a 1-bit flag, in case of '1', in-loop filter parameter information may be partially copied from a neighboring block, and in case of '0', in-loop filter parameter information not copied from a neighboring block may be selected.

If it is '1', flags can be transmitted in a preset order in order to know which information is to be partially copied from a neighboring block. For example, if the information on the starting position of the band to which filtering is applied means that the previous flag is '1', it is retrieved from the neighboring block at the same time, a flag indicating whether or not to obtain offset information for each remaining band. they can follow

If all four offsets are to be copied from neighboring blocks, a flag of 1-1-1-1 can be transmitted.

In addition, as shown in FIG. 11 (a), if all other bands except for the second band are to be copied from neighboring blocks, a flag of 1-0-1-1 can be transmitted, and in the case of the second band that is not copied, The corresponding offset can be transmitted separately.

As another example, it is assumed that the neighboring block performs in-loop filtering through the replacement value and the current block also performs in-loop filtering through the replacement value, and the parameter information of the neighboring block is as shown in FIG. 11(b).

Even in this case, all in-loop filter parameter information from a neighboring block may be copied or partial copy may be performed.

If the current block applies the in-loop filter through the replacement value, through the 1-bit flag indicating full copy of the parameter information of the neighboring block, if '1', the in-loop filter parameter information from the neighboring block is fully copied,' In the case of 0', it is possible to obtain parameter information of the current block from a neighboring block by selecting one of non-copying.

In addition, parameter information of the current block may be obtained from the neighboring block by partially copying the parameter information of the neighboring block. In this case, through a 1-bit flag, in case of '1', in-loop filter parameter information may be partially copied from a neighboring block, and in case of '0', in-loop filter parameter information not copied from a neighboring block may be selected.

If '1', flags may be transmitted in a preset order to know which information is to be partially copied from a neighboring block. If there are 4 bands to be filtered, a flag for copying or not can be transmitted for the start position of each band.

In addition, if all four bands want to copy the starting position from a neighboring block, a flag such as 1-1-1-1 can be transmitted.

If, as shown in FIG. 11 (b), if the starting positions of the remaining bands except for the last one are the same, the 1-1-1-0 flag can be transmitted and the starting position information for the remaining bands can be transmitted. can be transmitted through

If the starting positions of all four bands are copied from neighboring blocks, a flag indicating whether replacement values for the four bands are copied can be transmitted (1-1-1-1 or 1-0-1-1).

If, as shown in FIG. 11 (b), if only three copies of the band start position are copied and the start position is transmitted separately in the case of the last band, the replacement value copy flag is sent only for the three bands from which the band start position was copied and the last band is copied. In the case of a band, an alternative value may be transmitted separately (1-1-1-0 < band start position copy flag> ~~ 1-1-0 or 1-0-1 etc. < band whose band start position copy flag is 1 Copy the replacement value of the flag>).

The above-described neighboring blocks may be composed of various blocks, such as adjacent left, upper left, upper, upper right blocks, etc., or non-adjacent blocks and temporally adjacent co-located blocks. The neighboring block candidate group may be set according to rules determined by the encoder and decoder, or may be transmitted in units of pictures, slices, and the like. In addition, when two or more candidate groups exist, a flag of 1 bit or more may be transmitted.

Methods such as fixed-length encoding or variable-length encoding (Exp-golomb, truncated rice, etc.) can be applied to filtering-related parameters (offset or replacement value).

In addition, various starting positions for expressing related parameters can be set. If the left side of the corresponding band can be set as the center, or the center of the corresponding band can be set as the center. The range for expressing the related parameter can be limitedly set within the corresponding band, and it is also possible to set a range beyond the band.

In addition, it is possible to express related parameters based on standards such as an average value or a median value of pixels within a corresponding band. Relevant parameters can be expressed with absolute values and signs, and depending on the starting position, signs can be omitted and expressed only with absolute values.

When two or more bands are used, correction may be performed for all bands through offset or correction may be performed for all bands through replacement values.

Referring to FIG. 12, it is also possible to perform correction through an offset or replacement value adaptively according to a band. This information may be shared according to the same settings of the encoder and decoding, and may be transmitted in units of pictures, slices, or blocks.

Referring to FIG. 13, if it is determined that it is a special region of the image, correction through replacement values is added in addition to the edge offset and band offset in the existing SAO. Otherwise, edge offset and band offset as before can support only

For example, it is assumed that the video feature candidate selector selects 1. whether it is a text area or not from among several video feature candidates, and decides whether to add an additional mode to the existing SAO considering it.

Through the pixel classification unit 12, it is classified into several bands according to the luminance component. In order to determine whether the image analysis unit 13 is a character area or not, the number of subbands for which NZ_i is true in the corresponding area is more than 1 to less than k, and as shown in FIG. Assume that it is determined that the subband for which NZ_i is true) satisfies the condition of TH or greater (assuming that the difference between the luminance component of the text part and the luminance component of the background part is large).

In this case, if the analysis result that the block satisfies the condition is highly likely to actually be a character area comes from the mode addition decision unit 14, a mode called 'correction through substitution value' is applied to the existing edge offset and band offset is additionally executed, and if the corresponding block does not satisfy the condition, since it is unlikely to be a character area, SAO is supported only with the existing edge offset and band offset, excluding correction through replacement values.

When the optimal filter is determined by being executed in the addition mode, parameters related thereto are transmitted through the addition mode identification flag and the related parameter encoder 16.

The correction through the substitution value may be used like existing methods depending on the existing SAO, or may be configured and used with another in-loop filter.

14 is a block diagram for explaining a video encoding apparatus using image analysis according to an embodiment of the present invention.

Referring to FIG. 14, the video encoding apparatus according to an embodiment of the present invention includes a subtraction unit 110, a transform unit 120, a quantization unit 130, an inverse quantization unit 131, an inverse transform unit 121, an entropy It includes an encoder 140, an adder 150, an in-loop filter unit 160, a frame memory 170, an intra prediction unit 180, and a motion compensation unit 190.

The subtraction unit 110 generates a residual image between the current image and the predicted image by subtracting the predicted image generated by intra prediction or inter prediction from the received input image to be encoded (current image).

The transform unit 120 functions to transform the residual image generated by the subtraction unit 110 from a spatial domain to a frequency domain. Here, the transform unit 120 transforms the residual image into the frequency domain by using a technique for converting an image signal on the spatial axis into a frequency axis, such as Hadamard transform, discrete cosine transform, or discrete sine transform. can be converted to

The quantization unit 130 performs quantization on transformed data (frequency coefficients) provided from the transform unit 120 . That is, the quantization unit 130 divides the frequency coefficients, which are data converted by the transformation unit 120, by the quantization step-size and approximates them to calculate a quantization result value.

The entropy encoding unit 140 entropy-codes the quantization result value calculated by the quantization unit 130 to generate a bitstream. In addition, the entropy encoding unit 140 may entropy-encode the quantization result value calculated by the quantization unit 130 using CAVLC (Context-Adaptive Variable Length Coding) or CABAC (Context-Adaptive Binary Arithmetic Coding) techniques. In addition to the quantization result value, information necessary for decoding an image can be entropy-encoded.

The inverse quantization unit 131 inversely quantizes the quantization result value calculated by the quantization unit 130 . That is, the inverse quantization unit 131 restores a frequency domain value (frequency coefficient) from a quantization result value.

The inverse transform unit 121 restores the residual image by transforming the frequency domain value (frequency coefficient) received from the inverse quantization unit 131 from the frequency domain to the spatial domain, and the adder 150 restores the residual image for intra prediction or inter prediction. A reconstructed image of the input image is generated by adding the residual image reconstructed by the inverse transform unit 121 to the prediction image generated by the prediction image and stored in the frame memory 170 .

The intra predictor 180 performs intra prediction, and the motion compensator 190 compensates for motion vectors for inter prediction. Here, the intra prediction unit 180 and the motion compensation unit 190 may be collectively referred to as a prediction unit.

The prediction unit serves to generate a prediction block of the current block through intra or inter prediction. In the case of intra prediction, a prediction block can be constructed in a previously encoded or decoded region of the current picture using neighboring reference pixels using an extrapolation method, DC or planar mode, or an algorithm such as a block matching algorithm or template. An optimal prediction block may be determined in consideration of encoding cost.

Here, the encoded or decoded area may be configured in units of at least one block, may be divided in slice units, or may be configured as an area of a limited size in the direction in which encoding or decoding has been completed based on the current block. , may be composed of an area where encoding or decoding has been completed placed horizontally or vertically of the current block.

The parameters related to the optimal prediction mode determined at this time are block division information, prediction method (information on whether to use an extrapolation method or a method such as block matching), prediction direction, and motion information (which direction to use in case of extrapolation). In the case of block matching, etc., a motion vector for generating a prediction sample from a block at a movement distance from a specific position defined by the current block reference or the current slice start point reference or a specific method or from a pixel unit of an arbitrary shape. Information, information on which candidate to select among candidate lists for several motion vectors), composition of the candidate list (how many motion vectors to include as candidates, which motion vectors to put in the candidate group, etc.) information, motion search range (integer, 1/2, 1/4, 1/8, etc.), information for setting the derivation of a motion vector using a preset skip or merge, or other methods, and the like.

In particular, the intra predictor 180 according to an embodiment of the present invention improves prediction accuracy when motion search and compensation are performed for a target block to which horizontal and vertical inversions are applied using methods such as a block matching algorithm and a template. can make it For example, if an edge exists in the current block or if the current block is a screen content area, the intra prediction unit 180 may perform intra prediction on the current block by applying an inversion motion.

Whether or not to use a method such as a block matching algorithm and template to which inversion motion is applied may be set according to a previously agreed setting to the encoder and decoder, or may be transmitted in units of pictures, slices, or blocks. In addition, information on which inversion is to be supported can be shared according to settings agreed in advance to the encoder and the decoder. In addition, information on which inversion to support can be shared according to previously accumulated statistics (GOP units, picture units, etc.). Information related to inversion including the above-mentioned examples and other information may be included in parameters related to the optimal prediction mode.

The in-loop filter unit 160 performs filtering on the reconstructed image and may include a deblocking filter (DF) and a sample adaptive offset (SAO).

The in-loop filter unit 160 applies filtering to reduce an error between the restored image data and the original image data. An offset value for correcting the error between the restored image and the original image is calculated, and information about it is transmitted to the decoder. In the case of SAO, an offset in pixel units is added to a picture to which a deblocking filter is applied. SAO is subdivided into a case of correcting an error for an edge by subdividing pixel characteristics in a corresponding block (edge offset) and a case of correcting an error for a specific luminance component value band (band offset). For the purpose of efficiently calculating and transmitting offsets, four offsets may be transmitted for a luminance component and a chrominance component in a coding unit.

Furthermore, the in-loop filter unit 160 may configure various selection trees to perform filtering. That is, filtering may not be performed, correction may be performed through edge offset, correction may be performed through band offset, or correction may be performed through replacement values.

For example, the in-loop filter unit 160 may perform in-loop filtering on a reconstructed block obtained by reconstructing the current block according to an analysis result of image characteristics. That is, the in-loop filter unit 160 may perform filtering using a Sample Adaptive Offset (SAO) or a replacement value. For example, if an edge exists in the current block or if the current block is a screen content area, the in-loop filter unit 160 may perform filtering by applying a replacement value to a reconstructed block from which the current block is reconstructed. .

15 is a block diagram for explaining a video decoding apparatus using image analysis according to an embodiment of the present invention.

Referring to FIG. 15 , a video decoding apparatus according to an embodiment of the present invention includes an entropy decoding unit 210, an inverse quantization unit 220, an inverse transform unit 230, an adder 240, an in-loop filter unit 250 ), a frame memory 260, an intra prediction unit 270, and a motion compensation unit 280. Here, the intra prediction unit 270 and the motion compensation unit 280 may be collectively referred to as a prediction unit.

In particular, the intra prediction unit 270 may perform intra prediction on the current block by determining whether to apply an inversion motion based on a decoding method for the current block according to the result of analyzing the image characteristics. That is, the intra prediction unit 270 may perform intra prediction on the current block by determining whether to apply an inversion motion to intra prediction based on block matching or template matching based on the analysis result of the image characteristics.

For example, the intra prediction unit 270 may perform intra prediction on the current block by applying an inversion motion when an edge exists in the current block or when the current block is a screen content area.

The in-loop filter unit 250 may perform in-loop filtering on the current block based on a decoding method for the current block according to an analysis result of image characteristics. That is, the in-loop filter unit 250 may perform filtering using a Sample Adaptive Offset (SAO) or a replacement value. For example, if an edge exists in the current block or if the current block is a screen content area, the in-loop filter unit 250 may perform filtering by applying a replacement value to the current block.

Meanwhile, since each component of the video decoding apparatus can be understood in correspondence with the component of the video encoding apparatus of FIG. 14, a detailed description thereof will be omitted.

The video encoding/decoding device according to the present invention can be implemented as a computer-readable program or code on a computer-readable recording medium. A computer-readable recording medium includes all types of recording devices in which data that can be read by a computer system is stored. In addition, the computer-readable recording medium may be distributed to computer systems connected through a network to store and execute computer-readable programs or codes in a distributed manner.

In the video encoding/decoding method using image analysis according to the above-described embodiment of the present invention, intra prediction to which inverted motion is applied and in-loop filtering to which replacement values are applied to images having image characteristics different from natural images, such as screen content, are provided. Through this, it is possible to improve coding efficiency and reduce complexity.

Although the above has been described with reference to preferred embodiments of the present invention, those skilled in the art will variously modify and change the present invention within the scope not departing from the spirit and scope of the present invention described in the claims below. You will understand that it can be done.

11: image feature candidate selection unit 12: pixel classification unit
13: image analysis unit 14: mode addition determination unit
15: additional mode execution unit 16: additional mode identification flag and related parameter encoding unit
110: subtraction unit 120: conversion unit
121, 230: inverse transformation unit 130: quantization unit
131, 220: inverse quantization unit 140: entropy encoding unit
150, 240: addition unit 160, 250: in-loop filter unit
170, 260: frame memory 180, 270: intra prediction unit
190, 280: motion compensation unit 210: entropy decoding unit

Claims (6)

obtaining a prediction mode related parameter signaled from a bitstream, wherein the prediction mode related parameter includes at least one of a decoding method parameter and a candidate list configuration parameter;
determining whether a decoding method of a current block is a decoding method by block matching based on the decoding method parameter;
constructing a motion vector candidate list including a plurality of motion vector candidates of the current block based on the candidate list configuration parameter when the decoding method of the current block is the decoding method by block matching;
selecting one motion vector candidate from the configured motion vector candidate list;
generating a prediction block of the current block based on the selected motion vector candidate;
generating a reconstruction block of the current block by adding the generated prediction block to a residual block obtained from the bitstream; and
Applying an in-loop filter to a reconstruction block of the current block,
The selected motion vector candidate represents a position of a block matching the current block based on the current block;
The candidate list configuration parameter indicates the number of motion vector candidates constituting the motion vector candidate list;
A block matching the current block is a block that belongs to the same picture as the current block and has been decoded before the current block;
The in-loop filter is a sample adaptive offset filter including an edge offset filter and a band offset filter. According to claim 1,
The video decoding method of claim 1 , wherein the generation of the prediction block is performed by copying pixels of a block that matches the current block indicated by the selected motion vector candidate. determining whether an encoding method of a current block is a method by block matching;
generating a predicted block of the current block according to the block matching encoding method when the current block encoding method is the block matching encoding method;
generating a residual block of the current block by differentiating the generated prediction block from the original block of the current block; and
Encoding the residual block and prediction mode related parameters,
An in-loop filter is applied to a reconstructed block of the current block generated by adding the generated prediction block to the residual block;
The prediction block is generated using one motion vector candidate specified from among a plurality of motion vector candidates belonging to a motion vector candidate list;
The prediction mode related parameter includes at least one of an encoding method parameter or a candidate list configuration parameter,
The encoding method parameter indicates whether the encoding method of the current block is the encoding method by block matching;
The candidate list configuration parameter indicates the number of motion vector candidates constituting the motion vector candidate list;
The specified motion vector candidate indicates a position of a block matching the current block based on the current block;
A block matching the current block is a block that belongs to the same picture as the current block and has been decoded before the current block;
Wherein the in-loop filter is a sample adaptive offset filter including an edge offset filter and a band offset filter. According to claim 3,
The generation of the prediction block is performed by copying pixels of a block that matches the current block. A computer-readable recording medium storing a bitstream,
The bitstream includes prediction mode related parameters,
The parameter related to the prediction mode includes at least one of a decoding method parameter and a candidate list configuration parameter,
The decoding method parameter indicates whether the decoding method of the current block is a decoding method by block matching;
When the decoding method parameter indicates that the decoding method of the current block is the decoding method by block matching, the candidate list configuration parameter is used to construct a motion vector candidate list including a plurality of motion vector candidates of the current block, ,
A prediction block of the current block is generated based on a motion vector candidate selected from among a plurality of motion vector candidates belonging to the motion vector candidate list;
A reconstruction block of the current block is generated by adding the generated prediction block to a residual block obtained from the bitstream,
An in-loop filter is applied to a reconstruction block of the current block;
The selected motion vector candidate represents a position of a block matching the current block based on the current block;
The candidate list configuration parameter indicates the number of motion vector candidates constituting the motion vector candidate list;
A block matching the current block is a block that belongs to the same picture as the current block and has been decoded before the current block;
The in-loop filter is a sample adaptive offset filter including an edge offset filter and a band offset filter. According to claim 5,
The generation of the prediction block is performed by copying pixels of a block that matches the current block indicated by the selected motion vector candidate.

Images