KR20070090124A - Method and apparatus for decoding motion vectors - Google Patents

Abstract

A method and an apparatus for decoding motion vectors are provided to obtain predictive motion vectors of the same type even when a neighboring block has a different motion vector and reflect the obtained predictive motion vector to a process of obtaining a representative value(median) by using a motion vector prediction compensator so that encoding and decoding efficiency can be more improved. A method for decoding motion vectors comprises the following parts: selecting a neighboring block having a motion vector, which will be used as a predictive motion vector of a current block, according to a size and a location of the current block installed within a macro block; and selecting a motion vector of the selected neighboring block as the predictive motion vector of the current block. In the above first step, the motion vector of the neighboring block adjacent to an upper side of the current block is selected as the predictive motion vector of the current block if the size of the current block within the macro block is 8x16 and the current block is an upper block within the macro block.

Description

Motion vector decoding method and apparatus therefor {Method and apparatus for decoding motion vectors}

The present invention relates to encoding and decoding of a motion vector, and more particularly, to a method, a decoding, and an apparatus for encoding a motion vector calculated with reference to a plurality of pictures.

According to the MPEG-1 or MPEG-2 coding standard, in encoding a current picture, a P-picture encoding method for referring to a past picture and a B-picture encoding method for referencing both past and future pictures are employed. Based on this, motion compensation is performed. In addition, in order to improve encoding efficiency, the prediction vector is performed by using motion vectors of neighboring blocks so that the correlation between motion vectors between neighboring blocks is reflected instead of encoding the motion vector of the current block as it is.

The motion vector of neighboring blocks used for predictive encoding may not use the same reference picture as that of the current block. When such a case occurs, the conventional method has coped with a method of excluding a motion vector of a neighboring block having the same reference picture from a prediction encoding process. This is because, if a motion vector obtained based on another reference picture participates in predictive coding, coding efficiency may be reduced. However, if the motion vectors of neighboring blocks having other reference pictures are excluded from the predictive encoding process, the degree of reflecting the correlation of the motion vectors of the neighboring blocks may be lowered, and thus the encoding efficiency of the motion vectors may be reduced. This is even more the case in the case of a picture composed of blocks in which the reference picture is different from each other.

An object of the present invention is to provide a motion vector encoding method, a decoding method thereof, and an apparatus capable of performing more efficient predictive encoding even between motion vectors having different reference pictures.

Another object of the present invention is to provide a motion vector encoding method, a decoding method, and an apparatus, in which a reference picture readjusts at least some of different motion vectors to increase coding efficiency.

According to the present invention, a method of encoding a motion vector, the method comprising: (a) of a neighboring block having a motion vector of a different type from a motion vector of the current block among a plurality of neighboring blocks neighboring a current block; Predicting a motion vector of the same type as the motion vector to obtain a predicted motion vector; (b) calculating a representative value of motion vectors of the same type as the motion vector of the current block among the motion vectors of the neighboring blocks, including the obtained predicted motion vector; (c) calculating an error value between the calculated representative value and the motion vector of the current block; And (d) encoding the calculated error value.

According to another aspect of the present invention, in the apparatus for encoding a motion vector, the neighboring block having a motion vector of a different type from the motion vector of the current block among a plurality of neighboring blocks neighboring the current block may be generated. A prediction motion vector is obtained by predicting a motion vector of the same type as the motion vector of the current block, and then obtained from the motion vectors of the same type as the motion vector of the current block among the motion vectors of the neighboring blocks, including the obtained prediction motion vector. A motion vector predictor for calculating a representative value of the calculated value and calculating an error value between the calculated representative value and the motion vector of the current block; And a motion vector encoder for encoding the calculated error value.

When the motion vector of the current block is a forward motion vector MVf and the motion vector of the neighboring block is a backward motion vector MVb, the motion vector predictor calculates a forward motion vector MVf (prediction) that is the prediction motion vector by the following equation. It is desirable to.

MVf (prediction) = {distance of the past picture referenced by the current picture to which the neighboring block belongs / (distance of a future picture and the past picture referred to by the neighboring block-distance of a past picture to the current picture to which the neighboring block belongs)} * MVb

When the motion vector of the current block is backward motion vector MVb and the motion vector of the neighboring block is forward motion vector MVf, the motion vector predictor calculates backward motion vector MVb (prediction) that is the prediction motion vector by the following equation. It is preferable to calculate.

MVb (prediction) = {(distance of a future picture referenced by the neighboring block and the past picture-distance between a current picture to which the neighboring block belongs and a past picture referenced by the neighboring block) / distance of a current picture to which the neighboring block belongs to and a past picture referenced} * MVf

Preferably, the motion vector predictor obtains a predicted value PMV_X that is a representative value calculated by the following equation.

PVM_X = median (MVf_A, MVf_B, MVf_C) or median (MVb_A, MVb_B, MVb_C)

Here, MVf_A, MVf_B, MVf_C, MVb_A, MVb_B, and MVb_C are components of the motion vectors of the neighboring blocks, MVf_A is the forward motion vector of the left block, MVf_B is the forward motion vector of the upper block, and MVf_C is the right upper block. MVb_A is the forward motion vector of the left block. MVb_B refers to the backward motion vector of the upper block, and MVb_C refers to the backward motion vector of the right upper block. The prediction motion vector may be any one of them.

Meanwhile, according to another aspect of the present invention, the above object is a method for decoding a motion vector, comprising: (a) a neighboring block having a motion vector of a different type from the current block among a plurality of neighboring blocks neighboring the current block; Predicting a motion vector of the same type as the motion vector of the current block, and obtaining a predicted motion vector; (b) calculating a representative value of motion vectors of the same type as the current block among the motion vectors of the neighboring blocks, including the obtained predicted motion vector; And (c) adding the calculated representative value and the decoded error value to calculate a motion vector of the current block.

When the motion vector of the current block is a forward motion vector MVf and the motion vector of the neighboring block is a backward motion vector MVb, it is preferable that the forward motion vector MVf (prediction) that is the prediction motion vector is calculated by the following equation. .

MVf (prediction) = {time difference between the current picture to which the neighboring block belongs and a past picture referred to ((time difference between the future picture referred to by the neighboring block and the past picture minus the time difference between the current picture to which the neighboring block belongs and the reference past picture)} * MVb

When the motion vector of the current block is backward motion vector MVb and the motion vector of the neighboring block is forward motion vector MVf, it is preferable that the backward motion vector MVb (prediction) that is the prediction motion vector is calculated by the following equation. .

MVb (prediction) = {(time difference between future picture referenced by the neighboring block and the past picture minus time difference between the current picture to which the neighboring block belongs and the past picture referred to) / time difference between the current picture to which the neighboring block belongs and the past picture referenced} * MVf

When the motion vector of the current block is a forward motion vector MVf and the motion vector of the neighboring block is a backward motion vector MVb, it is preferable that the forward motion vector MVf (prediction) that is the prediction motion vector is calculated by the following equation.

MVf (prediction) = {distance of the past picture referenced by the current picture to which the neighboring block belongs / (distance of a future picture and the past picture referred to by the neighboring block-distance of a past picture to the current picture to which the neighboring block belongs)} * MVb

When the motion vector of the current block is backward motion vector MVb and the motion vector of the neighboring block is forward motion vector MVf, it is preferable that the backward motion vector MVb (prediction) that is the prediction motion vector is calculated by the following equation. .

MVb (prediction) = {(distance of a future picture referenced by the neighboring block and the past picture minus a distance between a current picture to which the neighboring block belongs and a past picture referenced by the neighboring block) / distance of a current picture to which the neighboring block belongs and the past picture referenced} * MVf

Step (b) is preferably a step of obtaining a prediction value PMV_X for decoding, which is a representative value calculated by the following equation.

PVM_X = median (MVf_A, MVf_B, MVf_C) or median (MVb_A, MVb_B, MVb_C)

Here, MVf_A, MVf_B, MVf_C, MVb_A, MVb_B, and MVb_C are components of the motion vectors of the neighboring blocks, MVf_A is the forward motion vector of the left block, MVf_B is the forward motion vector of the upper block, and MVf_C is the right upper block. MVb_A is the forward motion vector of the left block. MVb_B refers to the backward motion vector of the upper block, and MVb_C refers to the backward motion vector of the right upper block. The prediction motion vector may be any one of them.

According to another aspect of the present invention, an object of the present invention is to provide a motion vector decoder comprising: a motion vector decoder for decoding an error value; And a prediction motion vector is obtained by predicting a motion vector of the same type as the current block with respect to a neighboring block having a motion vector of a different type from the current block among a plurality of neighboring blocks neighboring the current block. Calculating a representative value for motion vectors of the same type as the motion vector of the current block among the motion vectors of the neighboring blocks, and calculating the motion vector of the current block by adding the calculated representative value and the error value. It is also achieved by an apparatus comprising a motion vector prediction compensator.

When the motion vector of the current block is a forward motion vector MVf and the motion vector of the neighboring block is a backward motion vector MVb, the motion vector prediction compensator calculates a forward motion vector MVf (prediction) that is the prediction motion vector by the following equation. It is preferable to calculate.

MVf (prediction) = {time difference between the current picture to which the neighboring block belongs and a past picture referred to ((time difference between the future picture referred to by the neighboring block and the past picture minus the time difference between the current picture to which the neighboring block belongs and the reference past picture)} * MVb

When the motion vector of the current block is backward motion vector MVb and the motion vector of the neighboring block is forward motion vector MVf, the motion vector prediction compensator calculates backward motion vector MVb (prediction) that is the prediction motion vector in the following equation. It is preferable to calculate by.

MVb (prediction) = {(time difference between future picture referenced by the neighboring block and the past picture minus time difference between the current picture to which the neighboring block belongs and the past picture referred to) / time difference between the current picture to which the neighboring block belongs and the past picture referenced} * MVf

When the motion vector of the current block is a forward motion vector MVf and the motion vector of the neighboring block is a backward motion vector MVb, the motion vector prediction compensator calculates a forward motion vector MVf (prediction) that is the prediction motion vector by the following equation. It is preferable to calculate.

MVf (prediction) = {distance of the past picture referenced by the current picture to which the neighboring block belongs / (distance of a future picture and the past picture referred to by the neighboring block-distance of a past picture to the current picture to which the neighboring block belongs) } * MVb

When the motion vector of the current block is backward motion vector MVb and the motion vector of the neighboring block is forward motion vector MVf, the motion vector prediction compensator calculates backward motion vector MVb (prediction) that is the prediction motion vector in the following equation. It is preferable to calculate by.

MVb (prediction) = {(distance of a future picture referenced by the neighboring block and the past picture minus a distance between a current picture to which the neighboring block belongs and a past picture referenced by the neighboring block) / distance of a current picture to which the neighboring block belongs and the past picture referenced} * MVf

On the other hand, the above object is also achieved by a computer readable information storage medium having recorded thereon program code for performing the above-described encoding method or decoding method.

In addition, according to another embodiment of the present invention, a method of decoding a motion vector includes selecting a neighboring block having a motion vector to be used as a prediction motion vector of the current block according to the size and position of a current block included in a macroblock; And selecting the motion vector of the selected neighboring block as a predictive motion vector of the current block, and selecting the neighboring block to be used as the predictive motion vector of the current block, wherein the size of the current block in the macroblock is 16x8. When the current block is an upper block in the macro block, the motion vector of the neighboring block adjacent to the upper side of the current block is selected as the predicted motion vector of the current block.

In addition, the apparatus for decoding a motion vector according to another embodiment of the present invention includes a motion vector decoder for decoding a difference value between the original motion vector of the current block and the predicted motion vector of the current block; And selecting a neighboring block having a motion vector to be used as a predictive motion vector of the current block according to the size and position of the current block included in a macroblock, and converting the motion vector of the selected neighboring block to the predictive motion vector of the current block. And a motion vector prediction compensator configured to add the selected predicted motion vector and the decoded difference value, wherein the motion vector prediction compensator has a size of a 16x8 current block in the macroblock and the current block is an upper block in the macroblock. In this case, the motion vector of the neighboring block adjacent to the upper side of the current block is selected as the predicted motion vector of the current block.

According to the present invention, in encoding and decoding a motion vector, a motion vector of the same type is predicted with respect to a neighboring block having a motion vector of a different type from the motion vector of the current block, and the intermediate value is obtained based on the intermediate value. Since the obtained difference value is encoded, the motion vector of the current block is obtained by predicting a motion vector of the same type with respect to a neighboring block having a motion vector of a different type from the decoded difference value and adding the intermediate value obtained based on the same. The coding efficiency and the decoding efficiency are increased.

That is, based on their representative value (middle value) of neighboring blocks in encoding or decoding the motion vector of the current block is based on the premise that the movements of neighboring blocks are similar. Accordingly, according to the motion vector encoding method and the decoding method of the present invention, in the case of a motion vector having a different conventional type, the motion of neighboring blocks is not properly reflected by being excluded from the representative value (middle value) calculation. Even in the case of having a motion vector, encoding and decoding efficiency are further improved by obtaining a predictive motion vector of the same type and reflecting the same in obtaining a representative value (middle value).

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

1 is a block diagram of a motion vector encoding apparatus according to a preferred embodiment of the present invention.

Referring to FIG. 1, the encoding apparatus 70 is an apparatus for encoding a motion vector, and includes a motion vector predictor 701 and a motion vector encoder 702. When motion vectors for a plurality of blocks are input, the motion vector predictor 701 may apply the neighboring block having only motion vectors of a different type from the motion vector of the current block among a plurality of neighboring blocks neighboring the current block. A prediction motion vector is obtained by predicting a motion vector of the same type as the motion vector of the current block, and then obtained from the motion vectors of the same type as the motion vector of the current block among the motion vectors of the neighboring blocks, including the obtained prediction motion vector. A representative value is calculated, and an error value between the calculated representative value and the motion vector of the current block is calculated. The motion vector encoder 702 encodes an error value output from the motion vector predictor 701. In the present embodiment, the motion vector encoder 702 encodes the difference value as an error value.

2 is a block diagram of an image data encoding apparatus including the motion vector encoding apparatus of FIG. 1, according to an exemplary embodiment of the present invention.

Referring to FIG. 2, the image data encoding apparatus is an encoding apparatus employing a multiple reference scheme, and includes an encoding controller 100, a first source encoder 200, and a second source encoder 700. do. Also, a source decoder 300, a memory 400, a motion compensator 500, and a motion predictor 600 are provided.

The input image data is composed of blocks obtained by dividing frames or frames input from a camera along a time axis into a predetermined size, as shown in FIG. 3. The frame includes a sequential scan frame obtained by the sequential scanning method, a field obtained by the interlaced scanning method or an interlaced scanning frame. Therefore, the video data described below means pictures of progressive scan frames, interlaced scan frames, fields, and block structures.

When image data is input, the encoding controller 100 determines a coding-type (intra coding / inter coding) according to whether to perform motion compensation on the input image according to the characteristics of the input image or a predetermined operation purpose desired by the user. To output the corresponding control signal to the first switch S1. When performing the motion compensation, the first switch S1 is closed because the image data inputted before or after is required, and when the motion compensation is not performed, the first switch S1 is not needed. S1 is opened. When the first switch S1 is closed, the difference image data obtained from the input image and the before or after image are input to the first source encoder 200, and when the first switch S1 is opened, only the input image is the first source encoder ( 200).

The first source encoder 200 quantizes transform coefficient values obtained by transform encoding the input image data according to a predetermined quantization step to obtain N × M data that is two-dimensional data composed of quantized transform coefficient values. . An example of the transform used may include a discrete cosine transform (DCT). Quantization is performed according to a predetermined quantization step.

Meanwhile, since the image data input and encoded by the first source encoder 200 may be used as reference data for motion compensation of the image data input after or before, the first source encoder by the source decoder 300. After inverse quantization and inverse transform encoding, which are inverse processes of 200, the memory is stored in the memory 400. In addition, if the data output from the source decoder 200 is difference image data, the encoding controller 100 closes the second switch S2 so that the difference image data output from the source decoder 300 is the motion compensator 500. And then stored in the memory 400.

The motion estimation unit 600 compares the input image data with data stored in the memory 400 to find the most similar data to the currently input data, and then compares the input image data with the input image data to obtain a motion vector MV ( Motion Vector) is output. The motion vector is obtained with reference to at least one picture. That is, it may be calculated with reference to a plurality of past and / or future pictures. When the motion vector is transferred to the memory 400, the memory 400 outputs the corresponding data to the motion compensation unit 500, and the motion compensation unit 500 currently encodes the image data based on the received data. Create and output the compensation value corresponding to

The second source encoder 700 receives information about the quantized transform coding coefficients output from the first source encoder 200 and the motion vector output from the motion predictor 600, and the encoding controller 100. Coding type information, quantization step information, and other information required for decoding are input and encoded to output a finally obtained bitstream.

The motion vector encoding apparatus 70 according to the present invention is provided in the second source encoder 700. Therefore, the motion vector encoding method according to the present invention is performed by the second source encoder 700. As in the MPEG-4 coding standard and the H.263 coding standard, the second source encoder 700 does not encode the motion vector of the current block as it is, but improves the coding efficiency by reflecting the correlation between the motion vectors. The prediction encoding is performed using the motion vectors of the two.

4 is a block diagram of a motion vector decoding apparatus according to a preferred embodiment of the present invention.

Referring to FIG. 4, the decoding apparatus is a device for decoding a motion vector according to the present invention, and includes a motion vector decoder 131 and a motion vector prediction compensator 132. The motion vector decoder 131 decodes the input encoded motion vector error value and restores the error value. In order to obtain the original motion vector from the reconstructed error value, it is necessary to obtain the predicted value used in the error value calculation and add it to the error value to finally find the original motion vector. To this end, the motion vector prediction compensator 132 receives motion vectors for a plurality of blocks and inputs them to a neighboring block having only motion vectors of a different type from the motion vector of the current block among a plurality of neighboring blocks neighboring the current block. Predicts a motion vector of the same type as the motion vector of the current block, and then obtains a predictive motion vector, and then, among the motion vectors of the neighboring blocks, including the obtained predicted motion vector, a motion vector of the same type as the motion vector of the current block. A representative value for the current block is calculated and the motion vector of the current block is calculated by adding the calculated representative value to the error value reconstructed by the motion vector decoder 131.

The motion vector decoding method according to the present invention is performed by the motion vector decoding apparatus 130 according to the present invention. As in the MPEG-4 coding standard and the H.263 coding standard, the motion vector decoding apparatus 130 does not decode the motion vector of the current block as it is, instead of decoding the motion vector of the current block in order to increase decoding efficiency by reflecting the association between the motion vectors. Predictive decoding is performed using motion vectors.

FIG. 5 is a block diagram of an image data decoding apparatus equipped with the motion vector decoding apparatus 130 of FIG. 4.

Referring to FIG. 5, a decoding apparatus includes a demux 110, a second source decoder 710, and a first source decoder 210 for demuxing a bitstream. A coding type information decoder 120 for decoding coding type information and a motion vector decoder 130 for performing motion vector decoding according to the present invention are provided.

The bitstream is demuxed into quantized transform coding coefficients, motion vector information, coding type information, and the like entropy coded by the demux 110. The second source decoder 710 entropy decodes the encoded transform coding coefficients and outputs the quantized transform coding coefficients. The first source decoder 210 source decodes the quantized transform coding coefficients. That is, the reverse process of the first source encoder 200 of FIG. 1 is performed. For example, if the first source encoder 200 performs the DCT, the first source decoder 210 performs an Inverse Discrete Cosine Transform (IDCT). Thus, the video data is restored. The reconstructed image data is stored in the memory unit 410 for motion compensation.

On the other hand, the coding type information decoder 120 detects the coding type and closes the third switch S30 in the case of an inter type requiring motion compensation, and then the motion compensator 510 to the data output from the first source decoder 210. Reconstructed image data is obtained by adding the motion compensation value output from The motion vector decoder 130 informs the position indicated by the motion vector obtained from the motion vector information, and the motion compensator 510 generates and outputs a motion compensation value from the reference image data indicated by the motion vector. Since the detailed operation of the motion vector decoder 130 is the same as that described with reference to FIG. 10, repeated description thereof will be omitted.

6 is a reference diagram for explaining a multiple reference method according to a preferred embodiment of the present invention.

Referring to FIG. 6, an I picture refers to an intra picture obtained without referring to another picture, and a B picture refers to a bi-predictive picture obtained by referring to two different pictures. P picture refers to a predictive picture obtained by referring to only an I picture. Arrows show pictures required for decoding, that is, dependency. The B2 picture depends on the I0 picture and the P4 picture, and the B1 picture depends on the I0 picture, the P4 picture, and the B2 picture. The B3 picture is dependent on the I0 picture, the P4 picture, the B1 picture, and the B2 picture. Thus, the display order is I0, B1, B2, B3, P4 ..., but the transmission order is I0, P4, B2, B1, B3 .... As described above, according to the encoding apparatus of FIG. 1, there is a B picture generated by referring to a plurality of pictures. In other words, the motion vector in the encoding method and the decoding method of the present invention includes at least one of a forward prediction method, a backward prediction method, a bi-directional prediction method, and a bidirectional prediction method including a direct prediction method. It is calculated in the way. In the bidirectional prediction scheme, both reference pictures may be both past pictures or both future pictures.

7 is a reference diagram showing motion compensation blocks that are a unit of motion compensation according to an exemplary embodiment of the present invention.

Referring to FIG. 7, a picture is composed of a plurality of motion compensation blocks. The motion compensation block is not only a 16x16 macroblock MB according to the present embodiment but also a 16x8 block obtained by dividing the macroblock in the horizontal direction, an 8x16 block obtained by dividing the macroblock in the vertical direction, 8x8 blocks obtained by dividing the macroblocks in the horizontal and vertical directions respectively, 8x4 blocks or 4x8 blocks obtained by dividing the macro blocks in the horizontal or vertical directions, respectively 4 Contains × 4 blocks.

The motion vector encoding method according to the present invention is based on a differential pulse coded modulation that encodes a representative value of neighboring blocks of the motion vector MV_X of the current block, and an error value with respect to a median in this embodiment. . That is, if the motion vectors of the left block A, the upper block B, and the right upper block C, which are neighboring blocks of the current block X, are MV_A, MV_B, and MV_C, respectively, the prediction vector PVM_X, which is a prediction value for the differential vector encoding, and the difference value using the same, The difference vector MVD_X is calculated by the following equation. The calculation is performed independently for each component value (x component and y component) of the vector. On the other hand, the left block A refers to the block to which the adjacent pixel to the left of the leftmost pixel of the first row belonging to the current block X belongs, and the upper block C refers to the pixel adjacent to the top of the leftmost pixel of the first row belonging to the current block X. The right upper block C refers to the block to which the adjacent pixel belongs to the right of the adjacent pixel above the rightmost pixel of the first row belonging to the current block X.

PMV_X = median (MV_A, MV_B, MV_C)

MVD_X = MV_X-PMV_X

The motion vector decoding method according to the present invention is based on a decoding method of obtaining a motion vector MV_X of a current block by adding a representative value of neighboring blocks of the current block, a median in this embodiment, and an error value MVD_X obtained by decoding. That is, if the motion vectors of the left block A, the upper block B, and the right upper block C, which are neighboring blocks for the current block X, are MV_A, MV_B, and MV_C, respectively, the prediction vector PVM_X, which is a prediction value for decoding, and the error value decoded therein. The motion vector MV_X obtained by adding MVD_X is calculated by the following equation. The calculation is performed independently for each component value (x component and y component) of the vector.

PMV_X = median (MV_A, MV_B, MV_C)

MV_X = MVD_X + PMV_X

The logic for processing an exception in obtaining a predicted motion vector according to the above-described encoding method and decoding method is as follows.

First, a case in which a specific block among neighboring blocks A, B, C, and D is located outside the picture (or slice) is processed as follows. A slice is an arbitrary unit of the minimum unit of a series of data strings having a start code and is composed of macro blocks. A slice does not exist across several pictures, but within one picture.

When both blocks A and D are located outside the picture (or slice), MV_A = MV_D = 0 and blocks A and D are regarded as having different reference pictures from the current block X.

(2) When blocks B, C, and D are all located outside the picture (or slice), MV_B = MV_C = MV_A, and when taking an intermediate value, MV_A must be the prediction vector PMV_X of the current block X.

If block C is located outside the picture (or slice), it is replaced with block D. Block D is an upper left block of the current block X and indicates a block to which an adjacent pixel is located above the leftmost pixel of the first row belonging to the current block X.

Second, when the neighboring blocks are intra coded, that is, if any of the neighboring blocks A, B, C, and D is intra coded, the motion vector of the block is set to 0 and it is assumed to have another reference picture.

Third, when the neighboring block has a reference picture different from the current block, the "predicted motion vector" is obtained according to the method proposed by the present invention, and then the obtained "predicted motion vector" is used to obtain an intermediate value. In other words, if the current block and the neighboring block participating in the prediction encoding have different types of motion vectors, the prediction motion vectors of the same type are obtained according to the method proposed by the present invention, and then the equations 1 and 2 Find the median by substituting the equation for the prediction vector.

8 is a reference diagram for describing a case of obtaining a predictive motion vector according to a preferred embodiment of the present invention.

Referring to FIG. 8, the current block X has a forward motion vector MVf_X, and neighboring blocks A and D have forward motion vectors MVf_A and MVf_D of the same type as the current block X. However, the neighboring block B has a different type of backward motion vector MVb_B from the current block X, and the neighboring block C has both a forward motion vector MVf_C and a backward motion vector MVb_C.

Median values according to Equations 1 and 2 described above should be obtained for motion vectors of the same type. The reason is that if a motion vector different from the current block participates in encoding and decoding, the encoding and decoding efficiency may be reduced. However, since the neighboring block B does not have the same type of motion vector, first, the prediction motion vector of the neighboring block B must be obtained. In other words, a forward motion vector of the same type as the current block X must be predicted.

9 shows an example of a method of obtaining a predicted motion vector.

Referring to FIG. 9, a forward motion vector MVf (prediction) or a backward motion vector MVb (prediction) as a prediction motion vector for a neighboring block may be predicted by the following equation.

MVb (forecast) = {(Distance_B-Distance_A) / Distance_B} * MVf

MVf (forecast) = {Distance_B / (Distance_B-Distance_A)} * MVb

Here, Distance_A is the picture distance (picture_distance) between the reference picture 1 and the reference picture 2 referred to by the current picture to which the corresponding neighboring block belongs, and Distance_B is the picture distance (picture_distance) between the current picture to which the corresponding neighboring block belongs and the reference picture 1. In this example, reference picture 1 is a past picture for the current picture and reference picture 2 is a future picture.

10 shows another example of a method of obtaining a predicted motion vector.

Referring to FIG. 10, a forward motion vector MVf (prediction) or a backward motion vector MVb (prediction) as a prediction motion vector for a neighboring block may be predicted by the following equation.

MVb (forecast) = {(TR_B-TR_A) / TR_B} * MVf

MVf (forecast) = {TR_B / (TR_B-TR_A)} * MVb

Here, TR_A is a time difference between a past picture and a future picture referenced by the current picture to which the corresponding neighboring block belongs, and TR_B is a time difference between a current picture and a past picture to which the corresponding neighboring block belongs. The time difference can be obtained from the counter value TR (Temporal Reference) to know the relative time order of the display between the pictures.

On the other hand, the motion vector encoding method according to the present invention has the characteristics of the B (Bi-predictive) picture, so the B picture will be described in more detail as follows.

The B picture does not necessarily have to be constrained to perform motion compensation with past and future pictures in time. Accordingly, the motion vector of the current block belonging to the B picture is independent of time or direction. In other words, since the B picture is obtained by referring to a plurality of pictures, it is highly likely that the current block and the neighboring block have different types of motion vectors. Nevertheless, the motion vector of the current block belonging to the B picture can be encoded and decoded regardless of the types of motion vectors of neighboring blocks by the motion vector encoding method and the decoding method according to the present invention.

Since the B picture has four types of motion compensation modes, there are four types of motion vectors. Forward motion vectors Forward MV (MVf), backward motion vectors Backward MV (MVb), bidirectional motion vectors Bi-predictive MV (MVf, MVb), direct motion vectors Direct MV. In the case of the direct motion vector, since the motion vector information is not transmitted to the receiver, it is not necessary to encode and decode. However, similar to the P picture, the remaining motion vectors are encoded or decoded using the prediction vector PMV obtained from the motion vectors of the neighboring blocks. .

As described above, since the motion vector of the current block belonging to the B picture is independent of time or direction, it does not necessarily mean that forward and backward are separated by a specific time or direction. Two-way also does not necessarily include the past and the future based on a specific point in time to distinguish a specific time or direction, or two-way must be included based on a specific point.

In calculating the prediction vector PMV, MVf and MVb use motion vectors of the same type as the neighboring block, respectively. That is, MVf uses only neighboring MVf and MVb uses only neighboring MVb in obtaining a prediction vector. However, since the bidirectional motion vectors have both, the necessary motion vectors are used. Therefore, when the prediction vector PMV is set in encoding or decoding the motion vector of the B picture, when the type of the neighboring motion vector is different, the same type of motion vector can be predicted by the method described with reference to FIGS. 9 and 10.

Referring back to FIG. 7, MVf_X = (1,2), MVf_A = (4,4), MVb_B = (-1, -1), MVf_C = (0,3), MVb_C = (-1, -2) In this case, according to the prior art, since MVb_B = (-1, -1) is different from the motion vector of the current block, it is excluded in obtaining an intermediate value and (0,0) is substituted instead. Therefore, the median value is (0, 3) and the difference value DMV_X is (1, -1). However, according to the present invention, instead of MVb_B = (-1, -1), the predicted (1, 1) is used to find the median. Therefore, the median value is (1,1) and the difference value DMV_X is (0,1).

On the other hand, the motion compensation block may have a rectangular shape as described above. In the case of the rectangular motion compensation block, since the prediction direction of the motion vector is known, the prediction vector predicted in another manner is used instead of using the median of the motion vectors of the three neighboring blocks as the prediction vector.

11 is a reference diagram for explaining a method of obtaining a prediction vector for a rectangular motion compensation block.

Referring to FIG. 11, when the motion compensation block is 8 × 16, the motion vector of the left block is used as the prediction vector of the block ①, and the motion vector of the right upper block is used as the prediction vector of the block ②. do.

When the motion compensation block is 16 × 8, the motion vector of the upper block among neighboring blocks is used as the prediction vector of block ③, and the motion vector of the left block among neighboring blocks is used as the prediction vector of block ④.

When the motion compensation block is 8 × 4, the prediction vector of block ⑤ is obtained by using the median value of the previous block as it is, and the motion vector of the left block among neighboring blocks is used as the prediction vector of block ⑥.

When the motion compensation block is 4x8, the prediction vector of block ⑦ is obtained by using the median value of the previous block as it is, and the motion vector of the upper block among neighboring blocks is used as the prediction vector of block ⑧.

However, when the reference picture of the motion vector set as the prediction vector is different from the reference picture of the current block, the existing intermediate value is used. The other exceptions are handled the same as in the above-described median prediction.

The motion vector encoding method and decoding method described above can be created by a computer program. Codes and code segments constituting the program can be easily inferred by a computer programmer in the art. In addition, the program is stored in a computer readable medium, and read and executed by a computer to implement a motion vector encoding method and a decoding method. The information storage medium includes a magnetic recording medium, an optical recording medium, and a carrier wave medium.

1 is a block diagram of a motion vector encoding apparatus according to a preferred embodiment of the present invention;

2 is a block diagram of an image data encoding apparatus including the motion vector encoding apparatus of FIG. 1, according to a preferred embodiment of the present invention;

3 is an example of an input image;

4 is a block diagram of a motion vector decoding apparatus according to a preferred embodiment of the present invention;

5 is a block diagram of an image data decoding apparatus including the motion vector decoding apparatus of FIG. 10, according to an embodiment of the present invention;

6 is a reference diagram for explaining a multiple reference scheme according to a preferred embodiment of the present invention;

7 is a reference diagram showing motion compensation blocks that become a unit of motion compensation according to an embodiment of the present invention;

8 is a reference diagram for explaining a case of obtaining a predictive motion vector according to a preferred embodiment of the present invention;

9 is an example of a method of obtaining a predicted motion vector;

10 is another example of a method of obtaining a predicted motion vector;

11 is a reference diagram for explaining a method of obtaining a prediction vector for a rectangular motion compensation block.

Claims (3)

In the method for decoding a motion vector, Selecting a neighboring block having a motion vector to be used as a predicted motion vector of the current block according to the size and position of a current block included in a macroblock; And Selecting the motion vector of the selected neighboring block as a predicted motion vector of the current block, The selecting of the neighboring block to be used as the prediction motion vector of the current block includes: when the size of the current block in the macro block is 16x8 and the current block is an upper block in the macro block, the motion of the neighboring block adjacent to the upper side of the current block; And a vector is selected as the predicted motion vector of the current block. The method of claim 1, Restoring a difference value between the original motion vector of the current block and the prediction motion vector; And Adding the reconstructed difference value and the predicted motion vector. In the motion vector decoding apparatus, A motion vector decoder for decoding a difference value between the original motion vector of the current block and the predicted motion vector of the current block; And Selecting a neighboring block having a motion vector to be used as a predictive motion vector of the current block according to the size and position of the current block included in a macroblock, and selecting a motion vector of the selected neighboring block as a predictive motion vector of the current block And a motion vector prediction compensator configured to add the selected prediction motion vector and the decoded difference value. When the size of the current block in the macro block is 16x8 and the current block is an upper block in the macro block, the motion vector prediction compensator converts a motion vector of a neighboring block adjacent to the upper side of the current block as a predicted motion vector of the current block. And a motion vector decoding device.

Images