KR100529783B1 - Prediction Direction Selection Method in Image Signal Prediction Coding - Google Patents

Abstract

The present invention selects a predictive block based on the difference between the neighboring blocks already encoded and transmitted and the DC coefficient of the current block, and increases the coding efficiency by performing AC coefficient prediction encoding in the predicted block direction. The present invention relates to a prediction direction selection method in predictive encoding.

The present invention calculates the difference between the DC blocks of the current blocks and the neighboring blocks that are already coded and transmitted, and selects the prediction direction of the block to be AC predictively encoded according to the difference. Since there is no need to transmit additional information, it is possible to reduce overhead and reduce additional information.

Description

Prediction Direction Selection Method in Image Signal Prediction Coding

The present invention eliminates the need to transmit additional information to the position of the prediction block by performing prediction encoding based on the difference between the neighboring blocks already encoded and transmitted and the DC coefficient of the current block, thereby reducing overhead and additional information. To increase the coding efficiency, and to perform the prediction encoding based on the difference between the actual data (neighbor block DC and the current block DC), thereby making prediction more accurate than using the slope of the neighboring block. It relates to a direction selection method.

In general, an intra-mode block (hereinafter, referred to as an intra block) that does not use temporal prediction in conventional standardization methods such as MPEG-1, JPEG, H.261, and H.263 using block-oriented encoding. Texture coding using 8 * 8 DCT (Discrete Cosine Transform; Discrete Cosine Transform).

The amount of data generated by the texture coating is a significant portion of the total data that needs to be sent to the decoder. Therefore, the efficient encoding of data generated from texture coding has a great influence on the encoding performance of the encoder.

In the above-described H.263, for an intra block that does not use time direction prediction, after performing 8 * 8 DCT, the DC coefficient is quantized to 8 bits, and transmitted by pulse code modulation (PCM). In MPEG2, in order to increase the coding efficiency of the DC coefficient, the DC values of neighboring blocks that have been transmitted as shown in FIG. 2 are converted to DPCM (Zig-Zag Scan (Luminance Block: Luminance Block)) and Sequential Scan (Color Block: Chrominance Block). Differential Pulse Code Modulation (Predictive Coding) and transmit (MPEG2 video recommendation, MPEG4 VM2.X and above, DC Prediction of DC coefficients in Intra Macroblocks).

When the image is divided into 16 * 16 pixels on the X and Y axes, that is, a macroblock size, if the X-axis coordinate of the macroblock to be encoded is 0, the first luminance component block in the macroblock is shown in FIG. As shown in FIG. 2, the predictive coded value is 128 and the color blocks are all 128.

As described above, 8 * 8DCT is performed in the encoder as shown in Equation (1), and IDCT (Inverse DCT) is performed in the decoder as in Equation (2). After DCT, the DC and AC coefficients are arranged as shown in FIG.

Herein, DC corresponds to F (0,0) of Equation (1) below, and AC corresponds to remaining coefficients except for F (0,0).

In the above, u, v, x, y = 0, 1, 2,... , 7, x, y = spatial position in the pixel domain, u, v = spatial position in the transform domain, u, x is the horizontal direction, and v, y is the vertical index. And C (u), C (v) = 1 / √2, for u, v = 0, otherwise 1 (see ITU-TRECOMMENDATION H.263 Annex A: Inverse transform accuracy sepification, 6.2.4: Inverse transform ). In addition, f (x, y) is a brightness value in an 8 * 8 block, and F (u, v) are DC and AC coefficients calculated using Equation (1).

In addition to the brightness (luminance, color), F (u, v) in Equation (2) is an intra-frame (I-Frame: First Frame of Image Sequence) having a relatively high spatial correlation between DC and values encoded by the encoder. In the case of I-VOP (Video Object Plane), which is the same concept as First VOP of VOP Sequence, encoding, considering only one neighboring block like MPEG2, it is difficult to expect high prediction encoding gain.

In the above description, the VOP is a unit for dividing an image into object units and encoding the respective images. For example, if there is a child and a cat in one image, the background may be divided into VOP0, a child in VOP1, and a cat in VOP2.

Therefore, encoding is possible for each VOP, which requires shape information to distinguish each object from each other. Shape information has different values for each VOP and distinguishes different VOPs according to this value.

In MPEG4, which is currently being standardized, the encoding efficiency of texture coding is improved by predictive encoding by using gradients between DC coefficients of blocks immediately above, to the left and to the left of the block to be encoded when AC / DC is encoded. Video Verification Model Version 4.0-3.4.4 Intra DC and AC Prediction, DC: DPCM, AC: PCM and DPCM.

However, in the conventional block-oriented encoding method, when the three neighboring blocks and the brightness value of the block to be encoded are different from each other, when the prediction encoding is performed using the DC coefficients of the neighboring blocks, the block to be encoded is used. It has caused a problem that the coding efficiency is lowered.

Accordingly, the present invention has been proposed to solve the prediction encoding gain degradation occurring in the conventional block-oriented encoding as described above, and an object of the present invention is to provide a difference between the DC coefficients of neighboring blocks that are already encoded and transmitted and the current block. The present invention provides a method for selecting a video signal prediction direction by eliminating the necessity of transmitting additional information to the position of a prediction block by performing predictive encoding, thereby reducing an overhead and increasing encoding efficiency.

Another object of the present invention is to select a prediction direction when predicting encoding of a video signal in which prediction encoding is performed based on a difference between actual data (neighboring block DC and current block DC) so that prediction is more accurate than using a slope of a neighboring block. To provide a method.

The method for achieving the object of the present invention is to calculate the difference between the DC block coefficient of the current block and neighboring blocks that are already coded and transmitted, and the prediction direction of the block to be AC predictively encoded according to the difference. Characterized in that the selection.

Hereinafter, described in detail with reference to the accompanying drawings of the present invention.

The coding system to which the present invention is applied is a block diagram of a MPEG-4 VOP coder determined primarily by the current international standard substructure as shown in FIG. This has a structure different from that of the H.261, H.263, MPEG-1, and MPEG-2 video encoding standards. In particular, the introduction of the concept of Shape Coder and VOP (Voideo Object Planes) shows the most significant difference. A VOP refers to an object at a point in time on an arbitrary shape of a content that can be accessed and edited by a user. The VOP must be encoded for each VOP to support content-based functionality.

When the VOP coder inputs a VOP for each object image formed by the VOP forming unit 20 to the MOTION ESTIMATION 31, the motion estimating unit 31 performs a macroblock unit from the applied VOP. The motion is estimated.

In addition, the motion information estimated by the motion estimator 31 is input to a motion compensation unit 32 to compensate for the motion. The VOP whose motion is compensated by the motion compensator 32 is input to the subtractor 33 together with the VOP formed by the VOP forming unit 11 to detect a difference value, and the difference value detected by the subtractor 33 is The internal information of the object is encoded by the object internal encoding unit 34 in units of sub blocks of the macro block.

For example, after the X and Y axes of the macroblock are subdivided into 8x8 subblocks having 8 pixels each with M / 2 x N / 2, the object internal information is encoded.

On the other hand, the VOP whose motion is compensated by the motion compensator 32 and the internal information of the object encoded by the object internal encoder 34 are input to the adder 35 and added, and the output signal of the adder 35 is transferred. The previous VOP, which is input to the VEV detecting unit 36 and is a VOP of the image immediately before the current image, is detected.

In addition, the previous VOP detected by the previous VOP detector 36 is input to the motion estimator 31 and the motion compensator 32 and used for motion estimation and motion compensation.

The VOP formed by the VOP forming unit 11 is input to a shape coding unit 37 to encode shape information.

Here, the output signal of the shape information encoder 37 varies depending on the field to which the VOP encoder is applied, and as shown by the dotted line, the output signal of the shape information encoder 37 moves the output signal of the shape information encoder 37. ), The motion compensation unit 32 and the object internal encoder 34 may be used to encode a motion estimation, motion compensation, and internal information of the object.

In addition, motion information estimated by the motion estimation unit 31, object internal information encoded by the object internal encoder 34, and shape information encoded by the shape information encoder 37 are applied to the multiplexer 38 to multiplex. Although not shown in the figure, the bitstream is transmitted to a multiplexer which multiplexes and outputs the outputs of the plurality of encoders.

In general, in a block-oriented coding scheme based on DCT, a macroblock (hereinafter abbreviated as MB) is configured as shown in FIG. 5.

In other words, MB consists of 16 horizontal and vertical pixels, and includes four luminance (Luminance) elements consisting of Block1 of Luminance (L1), Block2 of Luminance (L2), Block3 of Luminance (L3), and Block4 of Luminance (L4). It is roughly divided into two chromatic component blocks consisting of component blocks and Cb and Cr.

At this time, each pixel of C1 = Cb and C2 = Cr corresponds to a horizontally and vertically subsampled position of the luminance components, respectively, which is referred to as 4: 2: 0 image format and is illustrated in FIG. 3.

DCT is performed on brightness values within the six 8 * 8 blocks of MB described above, where one DC coefficient and 63 AC coefficients are calculated per block (see FIG. 1). A block using temporal prediction is referred to as an inter-mode block (hereinafter referred to as an inter block), and the inter block is an interframe coding method that improves coding efficiency by using a correlation between time frames between frames. Coding) refers to the size of one frame divided into 8 * 8 pixels, and is the number of unit pixels of texture coding using motion estimation (see ITU-T RECOMMENDATION H.263 Advanced Prediction Mode) and DCT. After the motion estimation, a compensation error is performed.

In general, an intra block refers to blocks constituting the intra frame in an interframe coding scheme, and when an inter frame is estimated by MB, a compensation error (estimated by SAD) is larger than a predetermined threshold. Four blocks in the MB correspond to this.

The present invention relates to predictive coding of AC coefficients quantized to 8 bits after texture coding with an intra block DCT, and 4 times for luminance (L1, L2, L3, L4) and 2 for color (C1, C2). A total of six times AC coefficient prediction encoding is performed.

In the case of color in the 4: 2: 0 video format, since Cb and Cr are present per MB, the unit of prediction coding is MB as shown in FIG. 2.

First, three neighboring blocks B1, B2, and B3 for the block B to be currently encoded are defined for encoding the DC coefficient.

As shown in FIG. 5, the six blocks L1, L2, L3, L4, C1, and C2 in any one MB are subjected to the zigzag-scan (L1, L2, L3, L4) and the sequential scan (C1, C2) of FIG. In order of encoding, for convenience, a block to be encoded is shown as B, an already encoded upper left block is B1, an upper block is B2, and a left block is B3.

In this case, the quantized DC coefficient of Block B is DC_B (Qunatized DC Block B), the block quantized DC coefficient of DC_B1 (Qunatized DC of Block B1), and the block B2 quantized DC coefficient of DC_B2 (Qunatized DC of Block B2). ), The quantized DC coefficient of block B3 is called DC_B3 (Qunatized DC of Block B3), and the predicted value for the prediction encoding of DC_B, which is the quantized DC coefficient of block B, is called DC_P (DC Predictor of a Qunatized Block B). When the predictive coded DC value for the current block B to be transmitted is DC_T (Predictive Differential Vaule of a Qunantized Block B to transmit on the Decoder), DC-T is defined as follows.

DC_T = (DC_P-DC_B)... … … … … … Formula (3)

Similarly to MPEG2, when the image is divided into 16 x 16 pixels on the X and Y axes, that is, the macroblock size, when the X-axis coordinate of the macroblock is 0 (the left boundary of the frame), the first luminance component block in the macroblock ( L1 and L3 of FIG. 4 set the predictive encoded value to 128, and the color blocks (C1 and C2 of FIG. 4) all set the predictive encoded value to 128. FIG.

In addition, when the Y-axis coordinate is 0 (right boundary of the frame), the first luminance component block (L1, L2 in FIG. 4) in the macroblock has a predictive coding value of 128 as shown in FIG. 2, and the color block (C1, All of C2) set the predictive coding value to 128.

Under these conditions, in the case of VOP Based Coding, DC prediction coding is as follows.

That is, when the blocks B1, B2, and B3 adjacent to the block B are all intra blocks, the absolute value of the result of subtracting the quantized DC coefficient DC_B2 of the block B2 from DC_B1, which is the quantized DC coefficient of the block B1 (| DC_B1-DC_B2 |) And the absolute value (| DC_B1-DC_B3 |) of the result of subtracting DC_B3, which is the quantized DC coefficient of block B3, from the quantized DC_B1 of block B1, compare (| DC_B1-DC_B2 | <| DC_B1-DC_B3 | DC predictive encoding is performed by using the prediction value DC_P for predictive encoding of DC_B of B as the quantized DC coefficient of block B3, otherwise, that is, if | DC_B1-DC_B2 || DC_B1-DC_B3 | DC prediction encoding is performed by using the prediction value DC_P for the prediction encoding as the quantized DC coefficient DC-B2 of the block B2.

Next, as shown in FIG. 5, when all neighboring blocks B2 and B3 of the block B to be encoded do not exist (convex that exists outside the shape information), the predicted value DC_P for predictive encoding of DC_B of the block B is 128. DC predictive encoding is performed.

If only block B3 among neighboring blocks B2 and B3 of the block B to be encoded exists, the DC prediction encoding is performed by using the prediction value DC_P for the prediction encoding of DC_B of the block B as the quantized BC coefficient DC_B3 of the block B3. Done.

Similarly, if only block B2 among the neighboring blocks B2 and B3 of the block B to be decoded exists, the DC predictive encoding is performed by using the prediction value DC_P for the prediction encoding of DC_B of the block B as the quantized DC coefficient of the block B2 (DC_B2). Will be performed.

In addition, if the neighboring blocks B2 and B3 of the block B exist but the block B1 does not exist, the prediction value DC_P for predictive encoding of the DC-B of the block B is set to the block B3 to mutually interact with the block B3 and the neighboring block. Predictive coding is performed with a relevant DC coefficient.

On the other hand, there is an exceptional situation that can occur when encoding using shape information in the above manner.

That is, when all of the blocks B2 and B3 adjacent to the block B to be encoded do not exist and only the block B1 exists, the prediction value DC_P for the prediction encoding of the DC_B of the block B is converted into the quantized DC coefficient DC_B1 of the block B1. In this case, DC prediction encoding is performed. If blocks B2 and B3 exist but block B1 does not exist, the prediction value DC_B1 for prediction encoding of block B1 is set to 128, and all neighboring blocks are intra blocks. Assuming this, the prediction value DC_P for predictive encoding of block B is obtained.

That is, the absolute value (| 128-DC_B2 |) of the result of subtracting the quantized DC coefficient DC_B2 of the block B2 from the prediction value 128 for the prediction encoding of the block B1 and the prediction value 128 for the prediction encoding of the block B1 Compare the absolute value (| 128-DC_B3 |) of the result of subtracting the quantized DC coefficient (DC_B3) of the block B3, | 128-DC_B2 | ＜ | 128-DC_B3 | In this case, DC predictive encoding is performed by using the predicted value DC_P for predictive encoding of DC_B of block B as the quantized DC coefficient DC_B3 of block B3. 128-DC_B2 | ＞ | 128-DC_B3 | In this case, DC prediction encoding is performed by using the prediction value DC_P for predictive encoding of DC_B of block B as the quantum DHK DC coefficient DC_B3 of block B2.

Next, other encoding methods may be considered for the above-mentioned possible exception situation.

If neither block B2 nor B3 exists and only block B1 exists, the prediction value DC_P for predictive encoding of DC_B of block B is added to the quantized DC coefficient DC_B1 of block B1 (DC_B1 + 128). The result is divided by 2 and the value of (DC_B1 + 128) / 2 decimal point is set to the DC coefficient by fnt (rounding up or down) the DC coefficient and DC prediction encoding is performed.

In addition, if only block B3 exists, the prediction value DC_P for the prediction encoding of DC_B of block B is added to the quantized DC coefficient DC_B3 of block B3 (DC_B3 + 128), and the result is divided by 2 (DC_B3). +128) / 2 The value below the decimal point is fnt (rounded up or down) to set the DC coefficient and DC prediction coding is performed.

If only convex B2 is present, the predicted value DC_P for predictive encoding of DC_B of block B is added to the quantized DC coefficient DC_B2 of block B2 (DC_B2 + 128), and the result value is divided by 2 (DC_B2). +128) / 2 The value below the decimal point is fnt (rounded up or down) to set the DC coefficient and DC prediction coding is performed.

In addition, when blocks B2 and B3 exist but block B1 does not exist, the predicted value DC_P for predictive encoding of DC_B of block B is converted into the quantized DC coefficient DC_B2 of block B2 to the quantized DC coefficient DC_B3 of block B3. ) Is added (DC_B2 + DC_B3), the result is divided by 2, and the value below (DC_B2 + DC_B3) / 2 decimal point is set to DC coefficient by fnt (rounding up or down) to perform DC prediction encoding.

The foregoing description has been made about the encoding method when the quantization slab sizes of the three blocks B1, B2, and B3 that are adjacent to the block B to be encoded are the same. When the quantization step sizes (hereinafter, referred to as Q_Step) of B2 and B3 are different, the DC values of the four blocks B, B1, B2, and B3 are normalized to the quantization step size of the corresponding block, and then the encoding method described above. Predictive coding of DC_B by applying.

Here, the quantization step size of each block is referred to as block B is Q_B, block 1 is Q_B1, block B2 is Q_B2, block B3 is Q_B3, and normalized DC is N_DC_B, N_DC_B1, N_DC_B2, N_DC_B3. DC normalization method is as follows.

N_DC_B = DC_B * Q_B

N_DC_B1 = DC_B1 * Q_B1

N_DC_B2 = DC_B2 * Q_B2

After normalizing to N_DC_B3 = DC_B3 * Q_B3, a prediction value DC_P for predictive encoding of block B to be encoded is obtained.

If DC_P is equal to N_DC_B1, DC_T, which is a predictive coded DC value for block B, is N_DC_B1 / Q_B-DC_B. When DC_P is equal to N_DC_B3, DC_T, which is a predictive coded DC value for block B, becomes N_DC_B3 / Q_B-DC_B to perform DC prediction coding.

The prediction direction of the AC coefficient according to the embodiment of the present invention is selected according to the difference between the DCs encoded and transmitted (hereinafter, DC means DC_P + DC_T, which is different from the above-described DC).

In order to describe a preferred embodiment of the present invention, two neighboring blocks A and B for the current block C are defined as shown in FIG. 6, and six blocks L1, L2, L3, in any one macroblock MB are defined. L4, C1, C2 are encoded according to the jig-zag scan (for L1, L2, L3, L4) and sequential scan (for C1, C2) of FIG. The upper block is shown as A, the left block is B.

In addition, the quantized AC / DC coefficient of A is A (i, j), the quantized AC / DC coefficient of B is B (i, j), and the quantized AC / DC coefficient of C is C (i, j), If the AC / DC prediction value of the prediction block P for C is P (i, j), and the coded AC / DC value for the block C transmitted from the encoder to the decoder is Cx (i, j), Cx (i, j) is

Cx (i, j) = P (i, j) -C (i, j) i, j = 0,1,... , 7... … … … … … Equation (4).

In the decoder, C (i, j) is reproduced (decoded) using the following equation (5).

C (i, j) = P (i, j) -Cx (i, j) i, j = 0,1,... , 7... … … … … Equation (5).

Herein, in the case of AC, the present invention relates only to predictive encoding of AC coefficients colored in block C of FIG. In addition, the main subject of the present invention is the DC coefficient [(A (0,0)) of the surrounding blocks (blocks A and B in FIG. 5) and the current block (block C in FIG. 5) already encoded and transmitted as described above. , B (0,0), C (0,0)] is used to encode the AC coefficient of the current block.

In this case, since the transmitted DC coefficients are used, it is not necessary to transmit additional information according to the AC prediction encoding. In addition, since the DC of the current block is used to select a prediction block for AC encoding (referring to one of A and B of FIG. 5 for predictive encoding), the DC of the neighboring block is more accurate. Efficient encoding can be performed, resulting in better efficiency than conventional methods.

The additional information refers to predefined information used to indicate the position of the prediction block in the bit stream transmitted to the decoder when the decoder does not know whether A or B is used as the prediction block for the current block. . For example, A may be defined as 0 and B may be 1.

First, when the blocks A, B, and C of FIG. 6 are all intra blocks, the following coding is performed using the already transmitted DC as described above.

If, | A (0,0) -C (0,0) | ＜ | If B (0,0) -C (0,0) |, block A is selected as the prediction block P. The predictively coded AC coefficient Cx to be transmitted to the decoder is generated by the following equation (6).

Cx (0, j) = C (0, j) -A (0, j) j = 1, 2,... , 7,

Cx (i, 0) = C (i, 0) i = 1,2,... , 7... … … … (6).

Unlike this | A (0,0) -C (0,0) | ＞ | B (0,0) -C (0,0) | If so, block B is selected as the prediction block P. The predictively coded AC coefficient Cx to be transmitted to the decoder is generated by the following equation (7).

Cx (i, 0) = C (i, 0) -B (i, 0) i, = 1,2,... , 7,

Cx (0, j) = C (0, j) j = 1, 2,... , 7... … … … (7).

I and j are the same indices as v and u of Equation (1), respectively, and when block A or B is outside of the object when shape information is used, AC of the corresponding block is replaced with 0, and the above coding is performed. To perform.

Next, when the quantization step sizes of blocks A, B, and C are different, only the prediction block is selected using the above-described method, and if the quantization step sizes of the selected prediction block and the current block are different, the block C to be encoded and the prediction block are different. When the quantization step sizes of (P) are different and block A is selected as the predictive block, AC coefficients A (0, 1), A (0, 2),. For A (0,7), when the quantization step size of block A is QA and the quantization step size of block C is QC, SA (0, j) * QA / QC = A (0, j) j = 1, 2,… , 7... … Scale with equation (8).

When the quantization step sizes of blocks A, B, and C are different, only the prediction block is selected using the above-described method, and the block C to be encoded is different from the quantization step size of the prediction block P, and the block B is a prediction block. Is selected, AC coefficients B (0.1), B (0.2),. For B (0.7), when the quantization step size of block A is QB and the quantization step size of block C is QC,

SB (i, 0) = B (i, 0) * QB / QC i = 1,2,... , 7... … Scale by equation (9).

Here, SA (0, j) and SB (i, 0) calculated in equations (8) and (9), respectively, are the scaled AC coefficients of blocks A and B.

Therefore, the equations (8) and (9) are used to scale the seven AC coefficients of the prediction block to obtain SA (0, j) and SB (i, 0). By substituting A (0, j) and B (i, 0), a predictive coded AC coefficient Cx to be transmitted to the decoder is obtained.

As described in detail above, the present invention has an effect of maximizing encoding efficiency by selecting an AC coefficient prediction encoding direction of a block to be encoded using a difference between DCs of blocks already encoded and transmitted.

1 is a position diagram of direct current (DC) and alternating current (AC) coefficients after a conventional discrete cosine conversion;

2 is a flowchart of a conventional DC prediction encoding.

3 is a 4: 2: 0 video format diagram;

4 is a configuration diagram of a MPEG-4 VOP video encoder primarily determined by the present International Standardization Organization;

5 is a block diagram of a macroblock MB applied to the present invention;

6 is an exemplary diagram of positions of neighboring blocks for AC prediction encoding in the present invention.

Claims (6)

A method of predictively encoding a video signal, the method comprising: calculating a difference between neighboring blocks already encoded and transmitted and a DC coefficient of a current block, and selecting a prediction direction of a block to be AC predictively encoded according to the difference A prediction direction selection method for encoding video signal prediction, characterized in that the. The method of claim 1, wherein if the blocks B (i, j) and A (i, j) neighboring the current block C (i, j) are all intra blocks in the AC coefficient prediction encoding, The absolute value of the result of subtracting A (0,0), the encoded DC coefficient of C, and C (0,0), the encoded DC coefficient of Block C, and B (0,0) and Block C, the encoded DC coefficients of Block B Compares the absolute value of the result of subtracting C (0,0), the encoded DC coefficient of | A (0,0) -C (0,0) | <B (0,0) -C (0,0) | Then, the block A is selected as the prediction block P, and a video signal characterized by generating the predictively coded AC coefficient Cx to be transmitted to the recuperator by the following equation (1). Prediction direction selection method in predictive encoding. Cx (0, j) = C (0, j) -A (0, j) j = 1, 2,... , 7, Cx (i, 0) = C (i, 0) I = 1,2,... , 7, … … … Formula (1) In the above, i, j is a spatial position in the transform domain. The method according to claim 2, wherein the blocks B (i, j) and A (i, j) neighboring the current block C (j, j) during the AC coefficient encoding are all intra blocks. The absolute value of the result of subtracting the coded DC coefficient A (0,0) and the coded DC coefficient C (0,0) of block C, and the coded DC coefficient B (0,0) and block C of block B Compares the absolute value of the result of subtracting C (0,0), the encoded DC coefficient of | A (0,0) -C (0,0) | ＞ B (0,0) -C (0,0) | Then, block B is selected as the prediction block P, and the predictive encoded AC coefficient Cx to be transmitted to the decoder is generated by the following equation (2). Prediction direction selection method in encoding. Cx (i, 0) = C (i, 0) -B (i, 0) i, = 1,2,... , 7, Cx (0, j) = C (0, j) j = 1, 2,... , 7, … … … Formula (2) In the above, i, j is a spatial position in the transform domain. 3. The block of claim 2, wherein the block DC to be encoded and the neighboring three blocks B1, B2, and B3 have different quantization step sizes. The absolute value of the result of subtracting C (0,0), which is the encoded DC coefficient of C, and B (0,0), which is the encoded DC coefficient of Block B, and C (0,0), which is the encoded DC coefficient of Block C If the prediction block is selected by comparing the absolute value of the subtracted result and the quantization step sizes of the selected prediction block and the current block are different, the AC coefficients of the selected prediction block must be scaled and transmitted to the decoder at the scaled value. A prediction direction selection method in predicting encoding of a video signal, characterized by calculating a predictive coded AC coefficient Cx. The method according to claim 4, wherein if block A is selected as the selected prediction block P, AC coefficients A (0, 1), A (0, 2),... For A (0,7), when the quantization step size of block A is QA and the quantization step size of block C is QC, after scaling to SA (0, j) = A (0, j) * QA / QC A method of selecting a prediction direction at the time of predictive encoding of image signal, characterized by generating the predictively encoded AC coefficient Cx to be transmitted by the following equation (4), Cx (0, j) = C (0, j) -SA (0, j) j = 1, 2,... , 7, Cx (i, 0) = C (i, 0) I =, 12,... , 7... … … … Formula (4) In the above, i, j is a spatial position in the transform domain. The method according to claim 4, wherein if block B is selected as the selected prediction block P, AC coefficients B (0, 1), B (0, 2),... For B (0,7), when the quantization step size of block B is QB and the quantization step size of block C is QC, after scaling to SB (0, J) = B (0, j) * QB / QC A prediction direction selection method in video signal prediction encoding, characterized by generating the predictive-coded AC coefficient Cx to be transmitted by the following equation (5). Cx (0, j) = C (0, j) -SB (0, j) j = 1, 2,... , 7, Cx (i, 0) = C (i, 0) I =, 1,2,... , 7. … … … … Formula (5) In the above, i, j is a spatial position in the transform domain.

Images