JP2007201675A - Moving picture encoder and moving picture encoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a moving picture encoder and a moving picture encoding method for calculating a proper weighted prediction table for the switching slice and reducing a processing amount in calculating the weighted prediction table. <P>SOLUTION: In encoding an input moving picture signal 10 as the SP slice, a switching section 180 is switched to a weighting prediction section 170 side a table calculation section 175 obtains a weight value and an offset value on the basis of a DCT factor converted by a DCT section 110 and the DCT factor converted by a DCT section 115 to set them to the weighting table and outputs them to the weighting prediction section 170. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a moving picture coding apparatus and a moving picture coding method for performing weighted prediction coding on a switching slice based on a weighted prediction table when coding an input moving picture signal in units of slices.

  In recent years, with the widespread use of the Internet and DVD recorders, digitally expressed moving images are becoming more and more common. As a technique for efficiently expressing a moving image with a small code amount, an inter-frame predictive encoding method typified by MPEG is widely used. However, in the inter-frame predictive encoding method up to MPEG4, there is a problem that noise is superimposed at the time of encoding because the amplitude fluctuation is not predicted in a moving image that is fading. In order to solve this problem, a new encoding method, H.264. In the H.264 / MPEG4-AVC (hereinafter abbreviated as AVC) encoding method, a weighted prediction method has been introduced (see, for example, Non-Patent Document 1). Amplitude fluctuations are predicted by this method, and high coding efficiency can be realized.

  In this AVC encoding system, SI (Swithing-I) slices are added to I / P / B slices for the purpose of smooth stream switching, error recovery, trick play, etc. during decoding. And switching slices called SP (Swithing-P) slices (see, for example, Non-Patent Document 1).

  By the way, in the weighted prediction in the AVC encoding method, an explicit mode calculated using an arbitrary coefficient and an implicit (Implicit) calculated using a coefficient corresponding to the distance between the picture to be encoded and the reference picture are used. There is a mode (see, for example, Non-Patent Document 1), but when performing weighted prediction in explicit mode, it is necessary to prepare a weighted prediction table by some method.

Therefore, in the conventional weighted prediction method, the weighted prediction table is set to a fixed value (see, for example, Patent Document 1). In another conventional example, the average value of the pixels of the entire slice for each slice for which weighted prediction is performed. (See, for example, Patent Document 2).
ISO / IEC 14496-10: 2004, Information technology-Coding of audio-visual objects-- Part 10: Advanced Video Coding JP 2004-7377 A JP 2004-7379 A

  However, in the conventional calculation method of the weighted prediction table described in Patent Document 1, since the weighted prediction table is a fixed value, weighted prediction cannot be performed using an appropriate weighted prediction table corresponding to an image. There is a problem.

  Further, in another conventional example described in Patent Document 2 described above, since the average value of luminance and color difference of the entire slice is calculated for each slice for which weighted prediction is performed, the average value is calculated for each pixel. There is a problem that the processing amount of weighted prediction increases.

  Accordingly, the present invention provides a moving picture coding apparatus and a moving picture coding method capable of calculating an appropriate weighted prediction table for a switching slice and reducing the processing amount when calculating the weighted prediction table. The purpose is to do.

  In order to solve the above-described problem, in the video encoding device of the present invention, when encoding an input video signal in units of slices, a video encoding device that performs weighted prediction encoding on a switching slice based on a weighted prediction table A table calculation unit that calculates the weighted prediction table based on a DCT coefficient obtained by DCT-transforming the input moving image signal, and the switching slice based on the weighted prediction table calculated by the table calculation unit. A weighted prediction unit that performs weighted prediction.

  In particular, in the image encoding device, the table calculation unit obtains a DC component value of the switching slice based on a DC component of the DCT coefficient, while an AC component of the switching slice based on an AC component of the DCT coefficient. A value is obtained, and the weight calculation unit obtains a weight value and an offset value when performing weight prediction based on the DC component value and the AC component value of the switching slice obtained by the table calculation unit, and performs weight prediction. .

  The video encoding method of the present invention is a video encoding method for performing weighted prediction encoding based on a weighted prediction table for a switching slice when encoding an input video signal in units of slices, The weighted prediction table is calculated based on DCT coefficients obtained by DCT-converting the input moving image signal, and weighted prediction is performed on the switching slice based on the calculated weighted prediction table.

  According to the moving picture coding apparatus and the moving picture coding method of the present invention, when performing weighted predictive coding based on the weighted prediction table for the switching slice, weighted prediction is performed based on the DCT coefficient obtained by DCT conversion of the input moving picture signal. Since a table is calculated and weighted prediction is performed on the switching slice based on the calculated weighted prediction table, an appropriate weighted prediction table can be calculated for the switching slice, and the weighted prediction table is calculated. Reduction of processing amount and parallel processing can be realized.

  Embodiments of a moving picture coding apparatus and a moving picture coding method according to the present invention will be described below. In the moving picture coding apparatus and the moving picture coding method according to the present embodiment, the AVC coding method is adopted as the coding method. As described above, the AVC encoding scheme includes I, P, B, SI, and SP slice types. Intra-coding can be performed in the I slice, intra-coding or inter-coding that refers to one picture in the P-slice, and intra-coding or inter-coding that refers to up to two pictures can be performed in the B slice. In addition, intra coding using the immediately preceding decoded frame as a reference image in the SI slice, intra coding using the immediately preceding decoded frame as a reference image, or inter coding using the immediately preceding decoded frame as a reference image may be performed in the SP slice. it can. These SI and SP slices are different from orthogonal transform coding of difference values performed in the I, P and B slices, and the difference values after orthogonal transform are coded.

  FIG. 1 is a block diagram showing an example of a configuration of a moving picture coding apparatus according to the present embodiment for coding a switching slice.

  In FIG. 1, the moving picture coding apparatus 100 according to the present embodiment includes DCT sections 110 and 115, quantization sections 120 and 125, variable length coding section 130, inverse quantization section 135, inverse DCT section 140, deblocking. Filter unit 145, memory unit 150, motion estimation (ME) unit 155, motion compensation (MC) unit 160, intra prediction unit 165, table calculation unit 175, weighted prediction unit 170, switching unit 180, subtractor 185, and adder The input video signal 10 is predictively encoded by the weighted predictive encoding method of the present embodiment, and the encoded bit stream 20 is output.

  FIG. 2 is a block diagram showing an example of the configuration of the moving picture decoding apparatus according to the present embodiment.

  2, the moving picture decoding apparatus 200 according to the present embodiment includes a variable length decoding unit 210, an inverse quantization unit 220, an inverse DCT unit 230, an intra prediction unit 240, an MC unit 250, a weighted prediction unit 260, It has a block filter unit 280, a memory unit 270, a switching unit 290, and an adder 295, which decodes the encoded bit stream 20 encoded by the video decoding device 100 shown in FIG. Is configured to output.

  Next, weighted prediction for switching slices in the video encoding device 100 and the video decoding device 200 will be described.

  That is, in the moving picture coding apparatus 100 according to the present embodiment, when a frame constituting the input moving picture signal 10 is coded as an SP slice, the switching destination of the switching unit 180 is a weighted prediction unit as shown in FIG. 170 side.

  Then, the table calculation unit 175 obtains a weight value and an offset value used for weighting based on the coefficient DCT transformed by the DCT unit 110 and the coefficient DCT transformed by the DCT unit 115, sets the weight value in the weighting table, and weights Output to the prediction unit 170.

  Then, the motion estimation (ME) unit 155 and the motion compensation (MC) unit 160 perform motion estimation and motion compensation on the local decoded image (reference image) of the previous frame stored in the memory unit 150, and a weighted prediction unit 170 performs weighted prediction based on the weight value and offset value set in the weighting table calculated by the table calculating unit 175, and outputs the weighted predicted inter prediction signal to the DCT unit 115.

  On the other hand, when a frame constituting the input moving image signal 10 is encoded as an SI slice, in the moving image encoding apparatus 100 according to the present embodiment, as shown in FIG. It becomes the prediction unit 165 side, and outputs the intra prediction signal to the DCT unit 115.

  The coefficients DCT transformed by the DCT units 110 and 115 are quantized by the quantizing units 120 and 125, respectively, and then subtracted by the subtracter 185. The difference is variable by the variable length coding unit 130. While being long-coded and output as the encoded bit stream 20, the inverse quantization unit 135 performs inverse quantization, the inverse DCT unit 140 performs inverse DCT, and the adder 190 passes the switching unit 180. The inter-prediction signal or the intra-prediction signal thus added is added, subjected to deblocking filter processing by the deblocking filter unit 145, and stored as a locally decoded image (reference image) in the memory unit 150.

  On the other hand, in the video decoding device 200 according to the present embodiment, when the encoded bit stream 20 is encoded as an SP slice, the switching destination of the switching unit 290 is on the weighted prediction unit 260 side as shown in FIG. The decoded signal inversely transformed by the inverse quantization unit 220 and the inverse DCT unit 230 and the decoded signal subjected to the weighted prediction at the time of motion compensation by the motion compensation (MC) unit 250 and the weighted prediction unit 260 are added. At 295, addition is performed, and the restored signal is output to the memory unit 270 via the deblock filter unit 280.

  On the other hand, in the moving picture decoding apparatus 200 according to the present embodiment, when the encoded bitstream 20 is encoded as an SI slice, the switching destination of the switching unit 290 is the intra prediction unit 240 as illustrated in FIG. The adder 295 adds the decoded signal inversely transformed by the inverse quantization unit 220 and the inverse DCT unit 230 and the decoded signal subjected to the intra prediction by the intra prediction unit 240, and the restored signal is decoded. The data is output to the memory unit 270 via the block filter unit 280.

  Here, it supplements about the inter prediction in the AVC encoding system in this Embodiment.

  In the AVC encoding method shown in the moving picture encoding apparatus 100 of the present embodiment, multi-reference using a plurality of reference pictures is adopted to improve encoding efficiency. Here, the reference picture is managed as a list, and is distinguished as L0 when mainly used for forward prediction, and as L1 when mainly used for backward prediction. Also, which reference picture is used is represented using a reference index.

  Next, weighted prediction by the weighted prediction unit 170 will be described.

  The weighted prediction in the weighted prediction unit 170 is defined as follows in Non-Patent Document 1 and the like described above. Specifically, the value before the weighted prediction of the position [x, y] is predPartL0C [x, y] in the case of L0 prediction, and predPartL1C [x, y] in the case of L1 prediction, and the position [x , y] after weighted prediction is predPartC [x, y] (luminance when C = L, blue difference when C = Cb, red difference when C = Cr). Further, BitDepthY and BitDepthC in the clip function are the bit depths of luminance and color difference.

  In particular,

  Then, in the case of one-way prediction (L0), the value after weighted prediction of position [x, y] is predPartC [x, y] as shown in the following formula (4) or formula (5).

  In the case of unidirectional prediction (L1), the value after weighted prediction of position [x, y] is predPartC [x, y] as shown in the following formula (6) or formula (7).

  In the case of bi-prediction, the value predPartC [x, y] after the weighted prediction at the position [x, y] is expressed by the following equation (8).

  Here, the weighted prediction table in the case of the weighted prediction in the explicit mode of the AVC encoding method is defined as the following equations (9) to (12). Here, the reference index indicating the reference picture for L0 prediction is refIdxL0, and the reference index indicating the reference picture for L1 prediction is refIdxL1.

  Here, in the case of luminance, logWD, weight values Wo and W1, and offset values Oo and O1 are expressed by the following equations (13) to (17).

  In the case of a color difference, logWD, weight values Wo and W1 and offset values Oo and O1 indicate Cb when iCbCr = 0 and Cr when iCbCr = 1, and the following equations (18) to (22) It becomes like this.

  However, at the time of bi-prediction, the weight values Wo and W1 comply with the restriction of the following equation (23).

  Next, an example of a method for reducing the processing amount when calculating the weight value and the offset value set in the weighting table by the table calculation unit 175 according to the present embodiment will be described.

  In the conventional table calculation unit, the weighted prediction table is calculated based on the input moving image signal 10 before DCT conversion. However, in the table calculation unit 175 of the present embodiment, as shown in FIG. , 115, the weight value and the offset value set in the weighted prediction table are calculated based on the DCT coefficients. Processing other than table calculation is the same as before.

  Hereinafter, although the calculation of the weighted prediction table is described as being performed on the luminance, the weighted prediction table can also be calculated for the color difference by the same process as the luminance.

  Here, a calculation method of a weighted prediction table used for conventional weighted prediction will be described. In the conventional calculation method, first, the DC component value dc_cur and the AC component value ac_cur, which are average values of the luminance values of the slice S of the encoding target frame, are obtained by the following equations (24) to (26).

  That is, with respect to the frame coordinates (x, y), the i-th block in the x direction and the j-th block in the y-direction are Bi and Bj, and the pixel value at the nth position (the top left is 0th) in the 4x4 block is p_n. When the number of pixels in S is N, the DC component value dc_cur and AC component value ac_cur of slice S are

It becomes. That is, the DC component value dc_cur is an average value of the pixel values p_n as shown in the above equation (24). Further, the AC component value ac_cur may use either the expression (25) or the expression (26), but the calculation amount of the expression (26) is larger than that of the expression (25).

  Next, the same calculation is performed for the reference indices refIdxL0WP and refIdxL1WP for the reference frame, and the DC and AC component values are set as dc_ref and ac_ref. From the above values, weight values Wo and W1 and offset values Oo and O1 for weighted prediction are calculated as in the following equations (27) to (30). The left bit shift by logWD is performed to approximate a value that is not bit-shifted to a power of 2.

  On the other hand, in the embodiment of the present invention, DCT coefficients that are outputs from the DCT units 110 and 115 are read into the table calculation unit 175, and a weighted prediction table used for weighted prediction is calculated based on the DCT coefficients. This is because the DCT conversion is performed before taking the difference value in the switching slice, and this DCT coefficient is used to represent the property of the image before conversion.

  FIG. 3 shows 4 × 4 DCT coefficients.

  As shown in FIG. 3, in the case of a 4 × 4 block, the DCT coefficient is the DC coefficient (DC component value) at the top left coefficient (q0) and the AC coefficient (AC component value) at the other coefficients (q1 to q15). . Here, the coefficient obtained by DCT-converting each pixel value of the 4 × 4 block is biased toward a low frequency term, and if it is discussed, the DC coefficient (DC component value) can also indicate each pixel value of the 4 × 4 block.

  Therefore, if the DC coefficient (DC component value) is added in the slice S and averaged by the number N of pixels in the slice S, the state of the pixel value in the slice S can be expressed as in the right side of the equation (24). Therefore, the table calculation unit 175 according to the present embodiment obtains the DC component value dc_cur by inputting the DC coefficient from the DCT units 110 and 115 by the following equation (31).

  Here, q0 represents a DC coefficient.

  Therefore, in the case of Expression (31), since the number of additions is only once for 16 pixels, the processing is performed in comparison with the conventional method of adding dc_cur 16 times for each pixel in the calculation expression (24) and averaging. The amount can be greatly reduced to 1/16.

  The coefficients other than the upper left coefficient q0 of the DCT coefficients are AC coefficients. The right side of the equation (25) or the equation (26) indicates the average deviation of each pixel value in the slice S with respect to dc_cur. Therefore, if the AC coefficient is added in the slice S, the equation (25) or the equation ( Similarly to the right side of (26), the state of the pixel value in the slice S can be expressed. Therefore, in the table calculation unit 175 of the present embodiment, the AC component value ac_cur is obtained by inputting the AC coefficient from the DCT units 110 and 115 by the following equation (32).

Here, q _n is the n-th DCT coefficient in the 4 × 4 block and indicates an AC coefficient. Compared to the conventional calculation formula (25) of the AC component value ac_cur, the calculation (32) according to the present embodiment eliminates the need for subtraction and absolute value calculation, thereby reducing the amount of calculation. .

  In addition, in the conventional equation (25), since the value of the DC component value dc_cur obtained by the equation (24) is used, the AC component value ac_cur cannot be calculated before the DC component value dc_cur is obtained. In the equation (32) according to the present embodiment, since the value of the DC component value dc_cur is not used, the DC component value dc_cur and the AC component value ac_cur are made independent by calculating the equations (31) and (32) simultaneously. And the processing can be parallelized.

  By the above method, the DC component value dc_cur and AC component value ac_cur of the switching slice S to be encoded, and the DC component value dc_ref and AC component value ac_ref of the reference frame are calculated, and from Equation (27) to Equation (30) Weight values Wo, W1 and offset values Oo, O1 are calculated. It should be noted that the weight values Wo and W1 are the ratio of the AC component values according to the equations (27) and (28), and therefore it is clear that it is not necessary to divide by the number of pixels N in the equation (32).

  Thus, according to the weighted prediction table calculation method of the present embodiment, when performing weighted prediction for a switching slice, the weighted prediction table is calculated based on the DCT coefficient after DCT conversion. An appropriate weighted prediction table can be prepared, and the amount of processing when calculating the weighted prediction table can be reduced and parallel processing can be realized.

  As a result, by reducing the processing amount, it is possible to improve the processing speed in terms of software, and it is possible to reduce power consumption in terms of hardware. Further, the parallel processing can improve the processing speed especially in hardware.

  In the above embodiment, as shown in the above equations (27) and (28), the AC component value ac_cur of the DCT coefficients is used for calculating the weight values Wo and W1 of the weighted prediction table. The AC component value ac_cur may be used for intra / inter determination. For example, in the SP slice, when the AC component value ac_cur of the current block is larger than the surrounding blocks, the block includes a lot of AC component values, so that the current block may be easily encoded as inter. . In addition, when DCT transform is performed in the P and B slices and the AC coefficient of the current block is larger than that of the surrounding blocks, the block includes a large number of AC component values, so that the current block is likely to be encoded as inter. It may be.

  Also, from the weighted prediction definition formula, it is obvious that when the weight values Wo and W1 are powers of 2, n bits are shifted to the right and n may be subtracted from logWD. Therefore, in order to reduce the code amount of the weighted prediction table, when the weight values Wo and W1 are close to the power of 2, the weighted prediction table may be approximated to the power of 2 and applied to the weighted prediction definition formula.

It is a block diagram which shows an example of a structure of the moving image encoder of embodiment. It is a block diagram which shows an example of a structure of the moving image decoding apparatus of embodiment. It is a figure which shows a 4x4 DCT coefficient.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Input moving image signal 20 Encoding bit stream 30 Output signal 100 Moving image encoding device 110,115 DCT part 120,125 Quantization part 130 Variable length encoding part 135 Inverse quantization part 140 DCT part 145 Deblock filter part 150 Memory unit 155 ME unit 160 MC unit 165 Intra prediction unit 170 Weighted prediction unit 175 Table calculation unit 180 Switching unit 185 Subtractor 190 Adder 200 Video decoding device 210 Variable length decoding unit 220 Inverse quantization unit 230 Inverse DCT unit 240 Intra Prediction Unit 250 MC Unit 260 Weighted Prediction Unit 270 Memory Unit 280 Deblock Filter Unit 290 Switcher 295 Adder

Claims (3)

When encoding an input video signal in units of slices, a video encoding device that performs weighted prediction encoding based on a weighted prediction table for a switching slice,
A table calculator that calculates the weighted prediction table based on DCT coefficients obtained by DCT-transforming the input moving image signal;
A weighted prediction unit that performs weighted prediction on the switching slice based on the weighted prediction table calculated by the table calculating unit;
A moving picture encoding apparatus having: The image encoding device according to claim 1,
The table calculation unit
Determining the DC component value of the switching slice based on the DC component of the DCT coefficient, and determining the AC component value of the switching slice based on the AC component of the DCT coefficient;
The weight calculation unit includes:
Obtaining a weight value and an offset value when performing weighted prediction based on the DC component value and the AC component value of the switching slice obtained by the table calculating unit, and performing weighted prediction;
A moving picture coding apparatus characterized by the above. A video encoding method for performing weighted prediction encoding based on a weighted prediction table for a switching slice when encoding an input video signal in units of slices,
The moving picture coding characterized in that the weighted prediction table is calculated based on DCT coefficients obtained by DCT-transforming the input moving picture signal, and weighted prediction is performed on the switching slice based on the calculated weighted prediction table. Method.