JP2008104216A - Video coder providing implicit coefficient prediction and scan adaptation for image coding and intra coding of video - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To decrease the bandwidth of signals, in coding video images. <P>SOLUTION: A predictive video coder performs gradient prediction, based on previous blocks (A, B and C) of image data. For a new block (X) of the image data, prediction determines the horizontal gradient and the vertical gradient, from a single block (B) located diagonally above the new block. Based on these gradients, the encoder predicts image information, based on the image processing of either an adjacent block in a horizontal direction, a vertically adjacent block or an adjacent block in the vertical direction, with respect to the new block. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  This application was granted by US Provisional Application No. 60 / 026,933 filed on September 20, 1996 and US Provisional Application No. 60 / 038,019 filed on February 14, 1997. It is based on priority.

  Various protocols for communication, storage and retrieval of video images are known. Protocols have been developed with a particular emphasis on always reducing the signal bandwidth. By reducing the signal bandwidth, the storage device can store more images and the communication system can send more images at a given communication speed. The reduction in signal bandwidth increases the overall capacity of the system that uses the signal.

  However, certain disadvantages can accompany the reduction in bandwidth. For example, certain known coding systems are lossy and introduce errors that can affect the perceptual quality of the decoded image. Other coding systems can achieve significant bandwidth reduction for certain types of images, but cannot achieve bandwidth reduction at all for other types of images. Therefore, careful consideration is required when selecting a coding scheme.

  Accordingly, there is a need in the art for an image coding scheme that reduces signal bandwidth without introducing perceptually significant errors.

  With the predictive coding scheme that predicts a new image data block from three image data blocks preceding the new block, the disadvantages of the prior art are greatly reduced. For this new block, the encoder examines the image data of blocks adjacent to the new block in the horizontal and vertical directions. The encoder positions the image data of each of the two adjacent blocks horizontally adjacent to the vertically adjacent block (above the new block diagonally), the image data of the third block Compare to. From these comparisons, horizontal and vertical slopes are determined. Based on the slope value, the encoder predicts the image data of a new block to be the image data of the block that is most similar to it in the horizontal or vertical direction. The encoder then determines the residual difference between the actual value of the image data and the predicted value of the image data for the new block and encodes the residue. The decoder performs inverse prediction that predicts the image data for the new block based on the horizontal and vertical gradients and adds the remainder to it to reconstruct the actual image data of the new block. There is no loss in this process.

  According to the present invention, video coding of coefficient data can be performed more efficiently by implicit gradient prediction and by scanning techniques that utilize the results obtained by the gradient prediction method. At present, the coding scheme of the present invention has been adopted in the MPEG-4 video verification model and is being considered for the MPEG-4 video standard.

  FIG. 1 (a) shows an encoder 100 constructed in accordance with the first embodiment of the present invention. An analog image signal is supplied to the encoder 100. The image signal is sampled using techniques known in the art and converted to a digital signal by an analog-to-digital (“A / D”) converter 110. The A / D converter 110 generates a digital image signal for a plurality of pixels of the image. Alternatively, the image signal can be supplied to the encoder as a digital image signal. In this case, the A / D converter 110 is deleted.

  The digital image signal is input to the processing circuit 120. The processing circuit 120 can perform a number of functions. Typically, the processing circuit 120 filters the image data and divides the image data into one luminance signal component and two color signal components. Additionally, the processing circuit 120 organizes the image data into data blocks. If the digital input signal represents information about a plurality of pixels in the scan direction, the digital input of the processing circuit 120 represents a pixel block, for example, blocks the data into an 8 x 8 pixel array of image data. be able to. The processing circuit 120 outputs image data on a macroblock basis. A macro block is typically composed of four luminance data blocks and two color data blocks. Processing circuit 120 also performs additional functions such as filtering to meet individual design criteria.

  The output of the processing circuit 120 is input to the transform circuit 130. Transform circuit 130 performs transformations of the image data, such as discrete cosine transform (“DCT”) coding or subband coding, from the pixel domain to the coefficient domain. The pixel block is transformed into a coefficient block with an equivalent size. Coefficients output by DCT coding generally include a single DC coefficient. The rest are AC coefficients that are partially non-zero. Similarly, the coefficients output by subband coding represent the characteristics of the image at various frequencies: Typically, many coefficients from subband coding are very small. The transform circuit 130 outputs a coefficient block.

  A quantizer 140 scales the signal generated by the transform circuit 130 according to a constant or variable scalar value (Qp). The quantizer 140 reduces the bandwidth of the image signal by reducing the number of quantization levels available to encode the signal. The quantization process is lossy. Numerous small coefficients input to the quantizer 140 are subdivided and truncated to zero. The scaled signal is output from the quantizer 140.

  The prediction circuit 150 performs gradient prediction analysis in order to predict the normalized DC coefficient of each block. Prediction circuit 150 can pass the scaled AC coefficients or alternatively can predict the AC coefficients of the block. In a preferred mode of operation, the prediction circuit 150 selects between an AC coefficient prediction mode or a pass mode. In this case, the prediction circuit 150 generates an AC prediction flag to identify the operation mode. The prediction circuit 150 outputs a DC residual signal, an AC signal (representing either an AC coefficient or an AC residual), and an AC prediction flag.

  A variable length coder 160 encodes the output of the prediction circuit 150. The variable length coder 160 is typically a Huffman encoder that performs run length coding on the scaled signal. The bitstream output from the variable length coder 160 can be transmitted, stored, or used for other purposes as known in the art.

  In the encoder 100, the prediction circuit 150 and the quantizer 140 perform functions independent of each other. Therefore, the operation order is not important. Although FIG. 1 illustrates the output of the quantizer 140 as an input to the prediction circuit 150, the order of the circuits can be reversed. The output of the prediction circuit 150 can be input to the quantizer 140.

  In FIG.1 (b), the decoder 200 performs the operation | movement which returns the above-mentioned encoding operation | movement. Variable length decoder 260 analyzes the bitstream using a complementary process to recover the scaled signal. When the variable length coder 160 of the encoder 100 uses a Huffman encoder, a Huffman decoder 260 is used.

  The restoration circuit 250 performs the same gradient analysis as that performed in the prediction circuit 150. The DC residual signal is identified and added to the predicted coefficient to obtain a DC coefficient. Optionally, the restoration circuit 250 can identify an AC prediction flag and interprets the AC information as either AC coefficient information or AC residual information based on the state of this flag. When AC residual information exists, the restoration circuit 250 adds the residual signal to the corresponding prediction signal to obtain an AC coefficient. The restoration circuit 250 outputs a coefficient signal.

  Scalar circuit 240 multiplies the recovered signal by the same scalar that was used as the base for division in quantization circuit 140. Of course, these coefficients divided to zero are not recovered.

  Inverse transform circuit 230 performs the inverse of the transform applied by transform circuit 130 of encoder 100. When a DCT transform is performed, an inverse DCT transform is applied. The same applies to the case of subband coding. The inverse transform circuit 230 transforms the coefficient information back into the pixel domain.

  The processing circuit 220 can combine the luminance and color signals to perform any function as desired in a particular application. The processing circuit 220 outputs digital signals of pixels that are ready for display. At this point, the signal is in a state suitable for display on a digital monitor. If it is necessary to adapt to a particular application, the signal can be converted by a digital-to-analog converter 210 for display on an analog display device.

  FIG. 2 illustrates the data structure as processed by the prediction circuit. The data output from the transform circuit represents a plurality of blocks organized into macroblocks. Each macroblock is typically occupied by four blocks representing the luminance component of the macroblock and two blocks representing the color component of the macroblock.

Each block represents the coefficient of the spatial region from which the block was derived. When a DCT transform is applied, the DC coefficient of the block's DC _x is provided at the block's origin in the upper left corner. AC coefficients are provided throughout the block, with the top coefficients being provided on the rows occupied by the DC coefficients in the horizontal direction and on the vertical rows occupied by the DC coefficients in the vertical direction. Yes.

  FIG. 3 shows a detailed block diagram of the prediction circuit 150. The quantizer 140 generates scaled DC and AC coefficients. The DC coefficients can be scaled (DC = DC / Qp, typically Qp = 8) and input to the DC coefficient predictor 300. The DC coefficient predictor performs gradient analysis.

For any block X, the DC coefficient predictor 300 is for block A that is horizontally adjacent to block X, block C that is vertically adjacent to block X shown in FIG. Data of block B, which is a block adjacent in the horizontal direction and adjacent to block A in the vertical direction, is maintained in the memory. The DC coefficient predictor compares the DC coefficient of block A (DC _A ) with the DC coefficient of block B (DC _B ). The difference between the DC coefficients of block A and block B is the vertical gradient. Similarly, the DC coefficient predictor 300 compares the DC coefficient (DC _B ) of block _B with the DC coefficient (DC _C ) of block C. The difference between the coefficients of block C and block B is the horizontal gradient.

The block associated with the highest gradient from block B is used as the basis for prediction. If the vertical gradient is greater than the horizontal gradient, it is expected that block A will have a high correlation with block X, so DC coefficient predictor 300 will use block A as the basis for prediction of block X. Use the horizontal prediction used. If the horizontal gradient is greater than the vertical gradient, then the DC coefficient predictor 300 utilizes vertical prediction, where it uses block C as the base for the prediction of block X. The DC coefficient predictor 300 outputs the DC coefficient (DC _A or DC _C ) of the block used for prediction to the subtractor 310. Similarly, the DC coefficient predictor 300 also generates a hor / vert signal 320 indicating whether horizontal prediction or vertical prediction is performed.

  The subtractor 310 subtracts the DC coefficient generated by the DC coefficient predictor 300 from the DC coefficient of the block X to obtain a DC residual signal for the block X. The DC residue is output from the prediction circuit 150 to the variable length coder 160.

  The process described above is used to predict the coefficients of a block within the image to be coded. However, when predicting coefficients at the start of a new row in the video object plane, the preceding block for prediction is the last block of the line that is up under normal processing. There is typically little correlation between these blocks.

  Assume that block Y in FIG. 2 is at the starting edge of the video object plane. In the scanning direction, no block is adjacent to block Y in the horizontal direction. Although the image data of the last block in the upper row is available to be used as a “horizontal adjacent” block, it is not used for prediction. Instead, the DC coefficient predictor 300 artificially sets the DC coefficient values for the horizontally adjacent block and the blocks above it to a half-intensity signal. If the DC coefficient is represented by an 8-bit word, the DC coefficient of these ghost blocks is set to 128. At this time, the DC coefficient predictor 300 performs gradient prediction according to the above-described process.

  As described above, the prediction circuit 150 can pass the AC coefficient without prediction. However, in a preferred embodiment, the prediction circuit 150 uses gradient analysis to predict AC coefficients.

When the prediction circuit 150 predicts the AC coefficient, only a part of the AC coefficient may indicate a high correlation between blocks. In the case of DCT transform coding and horizontal prediction, the only AC coefficients that are likely to show a sufficiently high correlation to be worthy of prediction analysis are those in the same tandem as the DC coefficients (within block A). With shadow). Therefore, for each AC coefficient (AC _X (0,1) to AC _X (0, n)) of the block X in the same column as the DC coefficient, the AC coefficient predictor 330 uses the block A (AC _A (0, n). 1) to AC _A (0, n)), the prediction corresponding to the AC coefficient arranged at the same place is generated. The predicted AC coefficient is subtracted from the actual AC coefficient of block X in a subtractor 340 to obtain an AC predicted residual signal.

In the case of DCT transform coding and vertical prediction, the only AC coefficient that has a high probability of exhibiting a correlation high enough for prediction analysis is one that is in the same row as the DC coefficient (with shadows in block C). ) For each AC coefficient (AC _X (1, 0) to AC _X (n, 0)) of block X in the same row as the DC coefficient, AC coefficient predictor 330 uses the AC coefficient of block C (AC _C (1 , 0) to AC _C (n, 0)). The predicted AC coefficient is subtracted from the actual AC coefficient of block X in a subtractor 340 to obtain an AC coefficient residual signal. The AC coefficient predictor is switched by a hor / vert signal 320 between the horizontal prediction mode and the vertical prediction mode. It is not necessary to perform gradient prediction of AC coefficients other than those described above.

  The AC coefficient correlation between blocks may occur, but it does not always occur. Therefore, AC coefficient prediction does not always lead to bandwidth efficiency. Thus, in a preferred embodiment, the prediction circuit 150 allows the selection of an operating mode between a mode in which AC coefficient prediction is performed and a second mode in which AC coefficient prediction is not performed. In the latter case, the AC coefficient from the transform circuit passes through the prediction circuit without change.

  When the residual is known, the AC prediction analyzer 350 is consumed by transmitting the macroblock AC residual signal and the bandwidth that should be consumed by transmitting the macroblock AC coefficients without prediction. Compare the expected bandwidth. Predictive analyzer 350 selects a transmission mode that consumes a relatively small bandwidth. Prediction analyzer 350 generates AC prediction flag signal 360 to display the selection.

Prediction is performed based on “similar species” blocks. When identifying blocks for prediction of luminance data blocks, only adjacent blocks of luminance data are considered. For prediction purposes, all intervening color data blocks are ignored. When predicting color block coefficients, only similar types of color signals are considered for prediction. When predicting data for a block of _Cr data that is one type of color signal, adjacent blocks of _Cr data are considered, but the intervening luminance block and the second type of color signal _CD data are It will be ignored. Similarly, when predicting the data for the C of the _D data block which is the second type of color signals is adjacent blocks of C _D data are considered, luminance blocks and C _r data intervening ignored The

  The prediction circuit 150 can output a DC residual signal, a signal representing either an AC coefficient or an AC residual, and an AC prediction flag signal.

  An inverse prediction operation is performed in the restoration circuit 250 shown in FIG. For all the blocks X, the DC coefficient predictor 400 includes the data of the adjacent block A preceding the block X, the data of the adjacent block C above the block X, and the block B preceding the block C that is a block above the block X. Data in memory. The DC coefficient predictor 400 compares the DC coefficient of block A with the DC coefficient of block B to determine the vertical gradient. Further, the DC coefficient predictor 400 compares the DC coefficient of block C with the DC coefficient of block B to determine the horizontal gradient. If the horizontal gradient is greater than the vertical gradient, the DC coefficient predictor 400 generates the DC coefficient of block C as a base for prediction. Otherwise, the DC coefficient predictor 400 generates DC coefficients for block A. The DC coefficient prediction 400 also generates a hor / vert signal 420 that identifies whether a horizontal or vertical prediction is being used.

  The recovery circuit 250 identifies the DC residual signal from the input bitstream. The adder 410 adds the DC residual to the DC coefficient generated by the DC coefficient predictor 400. The adder 410 outputs the DC coefficient of the block X.

  In the preferred embodiment, the recovery circuit 250 identifies the AC prediction flag 360 from the input bitstream. If the AC prediction flag 360 indicates that AC prediction was used, the restoration circuit identifies an AC residual signal from the input bitstream and operates the AC coefficient predictor 430. The hor / vert signal 420 from the DC coefficient predictor identifies whether block A or block C is used as the basis for the prediction. In response to this, the AC coefficient predictor 430 generates a signal corresponding to the AC coefficient of the block A or the block C in the same manner as the AC coefficient predictor 330 of the predictor 150. An adder 440 adds the predicted AC coefficient to the corresponding residual and outputs the reconstructed AC coefficient.

  If the AC prediction flag indicates that AC prediction was not used, the restoration circuit 250 identifies the AC coefficient signal from the bitstream. No arithmetic operations are required to restore the AC coefficient.

Sophistication of DC prediction can be achieved in the preferred embodiment by inducing a contribution to the DC coefficient of block X that is part of the perceptually dominant AC coefficient from the prediction block. For example, if block A is used as the prediction base, the predicted DC coefficients of block X are set as:
_{DC X = DC A + (4Q} p / 3) * (AC 02A -AC 01A / 4)
In the formula, Q _p is a quantity normalization factor, and AC _02A and AC _01A are AC coefficients of block A generated by the DCT transform.

Similarly, when block C is used as the basis for prediction, the predicted DC coefficient of block X may be set as follows:
_{DC X = DC C + (4Q} p / 3) * (AC 20C -AC 10C / 4)
In the formula, Q _p is a quantity normalization factor, and AC _20C and AC _10C are AC coefficients of the block C generated by the DCT transform.

  The prediction and restoration process described in this document is called an “implicit” method because no overhead signal is needed to identify which blocks are used for prediction. In operation, the coefficient values for blocks A, B and C are known by both encoder 100 and decoder 200. Thus, the decoder 200 can restore the prediction operation of the encoder without additional signal transmission. In embodiments where the prediction circuit did not select between modes of AC prediction, AC prediction and restoration is purely implicit. In the second embodiment, with the addition of the AC prediction flag, the prediction process is no longer purely implicit.

  The encoding / decoding operation of the prediction and reconstruction circuit can also be performed in software by a programmed microprocessor or digital signal processor.

  FIG. 5 illustrates the operation of the prediction circuit executed by software. The processor compares the DC coefficient of block A to the DC coefficient of block B to determine the vertical slope (step 1000). The processor similarly compares the DC coefficient of block C with the DC coefficient of block B to determine the horizontal slope (step 1010).

  The processor determines whether the vertical gradient is greater than the horizontal gradient (step 1020). If so, the processor defines the DC residual of block X as the actual DC coefficient of block X minus the DC coefficient of block A (step 1030). Otherwise, the processor defines the DC residual of block X as the actual DC coefficient of block X minus the DC coefficient of block C (step 1040).

If the processor also performs AC prediction, the processor operates as shown in FIG. Steps 1000-1040 are performed as described above with respect to FIG. If the vertical gradient is greater than the horizontal gradient, the AC coefficient from block A in the same column as the DC coefficient is used as a basis for predicting the corresponding AC coefficient of block X. Thus, for each such AC coefficients of the block _{X (AC X (0,1) ~AC} X (0, n)), the processor, corresponding in the block A from the actual AC coefficient in the block X AC The AC residual set to the coefficient (AC _A (0,1) to AC _X (0, n)) minus is calculated (step 1035).

  If block C is used as the basis for prediction, AC coefficients in the same row of DC coefficients may indicate a correlation between the blocks. Thus, for each AC coefficient AC (i) in the row of block X, the processor sets the residual (i) set to the actual AC coefficient in block X minus the corresponding AC coefficient in block C. Is calculated (step 1045).

  The processor also determines whether bandwidth savings have been achieved by predicting AC coefficients. Once all predictions have been made for a macroblock, the processor determines whether less bandwidth is achieved by the encoded coefficients or the residual (step 1050). If the residue has less bandwidth, the processor outputs the residue (step 1060). Otherwise, the processor outputs a coefficient (step 1070).

  In the preferred embodiment, additional bandwidth coefficients are obtained by constraining the scan direction of the variable length coder 160 for gradient prediction. The encoder scans the coefficient block to generate VLC coded run-level events. However, in natural images, there are often dominant preferred scanning directions. The present invention uses gradient prediction analysis to select one of three scan directions for performing run length coding.

  The first of the scanning directions is the horizontal alternating scanning shown in Table 1 above. The horizontal alternate search is used when the preferred scanning direction is the horizontal direction. Scanning starts from the origin, which is the DC residual position of the block. From the origin, the scan traverses three positions in the horizontal direction (0-3). From the fourth position, the scan jumps down to the first position in the second row. From the first position in the second row, the scan traverses three positions in the horizontal direction. The scan then jumps back to the first row of blocks and traverses the rest of the row. At the end of the first row, the scan jumps to the third row.

  The horizontal alternate scan traverses the next five passes (rows 3-8) in the same manner. From the first position in the i th row, the scan traverses three positions in the horizontal scan direction. Next, the scan jumps to (i-1) row and scans from the fifth position to the end of the row. At the end of the fifth pass, the scan jumps to the fifth position of the eighth row and traverses to the end of that row.

  The second of the scanning directions is the vertical alternating scanning shown in Table 2. Vertical alternating search is utilized when the preferred scanning direction is the vertical direction. Vertical alternating scanning is a complementary process to horizontal alternating scanning.

  From the origin, the scan traverses three positions in the vertical direction (0-3). From the fourth position, the scan jumps to the first position in the second column. From the first position in the second longitudinal row, the scan cuts through three positions vertically. Thereafter, the scan jumps back to the first vertical row of the block and traverses the remainder of the vertical row. At the end of the first column, the scan jumps to the third row.

  In the vertical alternate scanning, the next five passes (vertical rows 3 to 8) are vertically cut in the same manner. From the first position in the i-th row, the scan traverses three positions in the vertical scanning direction. The scan then jumps to (i-1) vertical line and scans from the fifth position to the end of the vertical line. At the end of the fifth pass, the scan jumps to the fifth position of the eighth column and traverses to the end of the column.

  The third scan direction is a conventional zigzag scan that is well known in the art.

  The variable length coder 160 selects the scan type according to the type of AC prediction performed. If the AC prediction flag 360 indicates that no AC prediction has been performed, the variable length coder 160 performs a conventional zigzag scan. When the AC prediction flag indicates that AC prediction is being performed, the variable length coder 160 performs conventional zigzag scanning. When the AC prediction flag indicates that AC prediction is being executed, the variable length coder 160 refers to the hor / vert signal 320 to determine whether horizontal prediction or vertical prediction is being used. In the case of horizontal prediction, scanning in the vertical diagonal direction is used. When vertical prediction is used, the variable length coder 160 utilizes horizontal-diagonal scanning.

  No additional overhead is required to determine the scan direction. The variable length decoder 260 determines the scanning direction. The AC prediction flag 360 and the halt / vert signal 420 are output from the restoration circuit 250. If the AC prediction flag 360 indicates that AC prediction has not been performed, the variable length decoder 260 assembles the coefficients according to the zigzag pattern. If the AC prediction flag indicates that AC prediction has been performed, the variable length decoder 260 assembles the residue based on the gradient prediction. The residue is assembled according to a vertical-diagonal scan in the case of horizontal prediction and by a horizontal-diagonal scan in the case of vertical prediction. In another embodiment, horizontal alternating and vertical alternating scans can be developed as shown in Tables 3 and 4 below, respectively.

  In yet another embodiment, horizontal alternating and vertical alternating scans can be developed as shown in Tables 5 and 6 below, respectively.

  The horizontal alternating scan in Table 5 starts at the origin, which is the DC residual position (position 0) of the block. From this origin, the scan steps in three positions in the horizontal direction (positions 0-3). The scan jumps to the first position in the second row below the origin (position 4). From position 4, the horizontal alternating scan steps one step in the horizontal direction (position 5), and then jumps to the first position (position 6) in the third row. The scan steps one position in the horizontal direction (position 7), returns to the second row at the third position (position 8), and steps one position across the row (position 9).

  From position 9, the horizontal alternating scan returns to the first row at the fifth position (position 10). The scan steps across to the end of the first row (positions 11-13). The scan returns to the second row at the end of the row (position 14) and scans horizontally across the row inward (positions 15-17) until the second row is complete. From position 17, the horizontal alternating scan returns to the third row at the fourth position (position 18), scans one step in the horizontal direction toward the origin (position 19), and jumps to the first position in the fourth row ( Position 20).

  From the first position in the fourth row, the horizontal alternating scan steps one position in the horizontal direction (position 21), then jumps to the first position in the fifth row (position 22), and again only one position horizontally. Step (position 23). The scan returns to the fourth row at the third position (position 24), scans across one step (position 25), and then returns to the third row at the fifth position (position 26). The scan steps horizontally across the third row to complete the row (positions 27-29).

  From the end of the third row, the horizontal alternating scan returns to the fourth row at the fifth position (position 30). The scan steps horizontally across the fourth row to complete the row (positions 31-33).

  From the end of the fourth row, the horizontal alternating scan returns to the fifth row at the third position (position 34). The scan steps one position in the horizontal direction (position 35) and then jumps to the first position in the sixth row (position 36). The scan steps across one position (position 37) and then jumps to the first position in the seventh row (position 38). The horizontal alternating scan steps across one position (position 39) and then returns to the sixth row at the third position (position 40). The scan steps across one position (position 41) and then returns to the fifth row at the fifth position (position 42). The horizontal alternating scan steps horizontally across the fifth row to complete this row (positions 43-45).

  From the end of the fifth row, the horizontal alternating scan returns to the sixth row at the fifth position (position 46) and steps across the horizontal to complete the row (positions 47-49).

  From the end of the sixth row, the horizontal alternating scan returns to the third position of the seventh row (position 50). The scan steps one position in the horizontal direction (position 51) and then jumps to the first position in the eighth row (position 52). The scan is stepped 3 positions in the horizontal direction (positions 53-55) and then returned to the seventh row at the fifth position (position 56). The horizontal alternating scan completes the row across the horizontal (positions 57-59). From the end of the seventh row, the scan jumps to the fifth position of the eighth row (position 60) and steps across the horizontal to complete the row (positions 61-63).

  The vertical alternating scan in Table 6 begins at the origin, which is the position of the DC residual of the block (position 0). The scan steps in three places in the vertical direction (positions 0-3). The scanning is longitudinally cut from the origin and jumps to the first position in the second longitudinal row (position 4). From position 4, the vertical alternating scan steps one step in the vertical direction (position 5) and then jumps to the first position of the third longitudinal row (position 6). The scan steps one position in the vertical direction (position 7), then returns to the second longitudinal row at the third position (position 8) and steps one position in the vertical direction (position 9).

  From position 9, the vertical alternating scan returns to the fifth position of the first longitudinal row (position 10). Scanning steps to the end of the first column (positions 11-13). The scan returns to the second vertical line at the end of the vertical line (position 14) and scans vertically through the vertical line toward the interior of the vertical line (positions 15-17) until the second vertical line is completed. From position 17, the vertical alternating scan returns to the third vertical line and the fourth position (position 18), scans one stage in the vertical direction toward the top of the vertical line (position 19), and the fourth vertical line Jump to position 1 (position 20).

  From the first position in the fourth column, the vertical alternating scan steps one position in the vertical direction (position 21), then jumps to the first position in the fifth column (position 22) and again in the vertical direction Step one position (position 23). Scanning returns to the fourth vertical row (position 24) at the third position, steps one step in the vertical direction (position 25), and then returns to the third vertical row at the fifth position (position 26). The scan is stepped vertically through the third column to complete the column (positions 27-29).

  From the end of the third column, the vertical alternating scan returns to the fourth column at the fifth position (position 30). The scan is stepped vertically through the fourth column to complete the column (positions 31-33).

  From the end of the fourth vertical row, the vertical alternating scan returns to the fifth vertical row at the third position (position 34). The scan steps one position in the vertical direction (position 35) and then jumps to the first position in the sixth column (position 36). The scan is stepped vertically one position (position 37) and then jumps to the first position in the seventh column (position 38). The vertical alternate scan steps one position in the vertical direction (position 39) and then returns to the sixth longitudinal row at the third position (position 40). Scanning is stepped one position in the vertical direction (position 41), then returned to the fifth position (position 42) of the fifth vertical row, stepped vertically through the fifth vertical row, and the vertical row Is completed (positions 43 to 45).

  From the end of the fifth vertical line, the vertical alternate scanning returns to the fifth position of the sixth vertical line (position 46), and the sixth vertical line is stepped vertically in the vertical direction to complete the vertical line ( Positions 47-49).

  From the end of the sixth column, the vertical alternating scan returns to the third position of the seventh column (position 50). The scan steps one position in the vertical direction (position 51) and then jumps to the first position of the eighth vertical line (position 52). The scan steps three positions vertically (positions 53-55) and then returns to the fifth position in the seventh column. The scan steps vertically through the seventh column to complete the column (positions 57-59). From the end of the seventh vertical row, the vertical alternating scan jumps to the fifth position of the eighth vertical row (position 60) and steps vertically through the eighth vertical row to complete the vertical row (position 61). ~ 63).

(A) is a schematic diagram of an encoder according to an embodiment of the present invention, and (b) is a schematic diagram of a decoder according to an embodiment of the present invention. It is a figure which shows an example of the image data processed by this invention. It is a prediction block diagram of FIG. 2 is a block diagram of the restoration circuit of FIG. It is a flowchart of the prediction circuit implement | achieved in software. 4 is a flowchart of a second embodiment of a prediction circuit implemented in software.

Explanation of symbols

  100 encoder, 110 analog-digital converter, 120 processing circuit, 130 transform circuit, 140 quantizer, 150 prediction circuit, 160 variable length coder, 200 decoder, 210 digital-analog converter, 220 processing circuit, 230 Inverse transform circuit, 240 scalar circuit, 250 restoration circuit, 320 adaptive DC coefficient predictor, 330 adaptive AC coefficient predictor, 350 AC prediction analyzer, 400 DC coefficient predictor, 430 AC coefficient predictor.

Claims (10)

Formatting the video signal into a plurality of vertically and horizontally adjacent image data blocks;
For each block of new image data,
Determining a horizontal gradient between a first image data block vertically adjacent to a block of new image data and a second image data block horizontally adjacent to the first block;
Determining a vertical gradient between a third image data block and a second image data block horizontally adjacent to the new block of image data;
Predicting image data for a new block of image data based on horizontal and vertical gradients;
Generating residual image data for the block of new image data from the predicted image data and the new image data;
A video signal encoding method comprising: The prediction step is further
Predicting image data for a new block of image data based on the image data of the first block if the vertical gradient is less than the horizontal gradient; and
Predicting image data for a new block of image data based on the image data of the third block if the horizontal gradient is less than the vertical gradient;
The video signal encoding method according to claim 1, further comprising: Each of the image data of the first, second and third blocks and the new image data includes DC coefficients of a discrete cosine transform;
Determining a horizontal gradient determines a horizontal gradient based on the DC coefficients of the first and second image data blocks;
Determining a vertical gradient determines a vertical gradient based on the DC coefficients of the second and third image data blocks;
The method of claim 1. The image data of the first, second and third blocks and the new image data each further include a plurality of AC coefficients of the discrete cosine transform;
The step of predicting further includes the step of predicting the AC coefficients of the block of new image data based on the horizontal and vertical gradients;
The generating step further includes, for each predicted AC coefficient, generating an AC residual signal based on the actual AC coefficient of the new block of image data and the predicted AC coefficient;
The method of claim 3. In a method for decoding a coded signal of a block of new image data,
A first image data block vertically adjacent to a new image data block, a second image data block horizontally adjacent to the first block, and a new image data block horizontally adjacent to the vertical direction Identifying a third image data block adjacent to the second block;
Determining a horizontal gradient between the first block and the second block;
Determining a vertical gradient between the third block and the second block;
Predicting image data for a new block of image data based on horizontal and vertical gradients;
Identifying a residual image signal from the coded signal; and
Generating an image signal for a new block of image data from the predicted image data and the residual image signal;
Comprising a method. The prediction step is further
Predicting image data for a new block of image data based on the image data of the first block if the vertical gradient is less than the horizontal gradient; and
Predicting image data for a new block of image data based on the image data of the third block if the horizontal gradient is less than the vertical gradient;
6. The method of decoding a video signal according to claim 5, wherein: Each of the first, second and third blocks includes DC coefficients of a discrete cosine transform;
Determining a horizontal gradient determines a horizontal gradient based on the DC coefficients of the first and second image data blocks;
Determining a vertical gradient determines a vertical gradient based on the DC coefficients of the second and third image data blocks;
The video signal decoding method according to claim 5. Receiving an input video signal;
Formatting the video signal into vertically and horizontally adjacent blocks of digital image data;
Transforming the block into a block of coefficients representing digital image data; and
About the new coefficient block;
Determining a horizontal gradient between a first coefficient block vertically adjacent to the new block and a second coefficient block horizontally adjacent to the first block;
Determining a vertical gradient between a third coefficient block and a second block horizontally adjacent to the new block;
Predicting coefficients for a new block based on horizontal and vertical gradients;
Generating residual coefficients for a new block of new image data from the predicted coefficients and the new coefficients;
A bitstream generated according to the method consisting of: In an encoder for encoding video information represented as a plurality of vertically adjacent and horizontally adjacent image data blocks each containing a DC coefficient,
New image data from the DC coefficients of the first block vertically adjacent to the new block, the second block horizontally adjacent to the first block, and the third block horizontally adjacent to the new block A DC coefficient predictor that predicts DC coefficients for the block; and
A subtractor for generating a DC residual signal for the new block representing the difference between the DC coefficient of the new block and the DC coefficient predicted for the new block;
An encoder comprising: In a video decoder for generating video information from received encoded signals, each represented as a plurality of vertically adjacent and horizontally adjacent encoded data blocks each containing DC residual information,
New image data from the DC coefficients of the first block vertically adjacent to the new block, the second block horizontally adjacent to the first block, and the third block horizontally adjacent to the new block A DC coefficient predictor for predicting DC coefficients for a block of; and
An adder for generating a DC coefficient signal based on the received DC residual information and the predicted DC coefficient for the new block;
A decoder comprising:

Images