JP2009260494A - Image coding apparatus and its control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform prediction coding in a frame by reducing operation resources required for prediction mode selection while achieving high throughput processing performance. <P>SOLUTION: A mode 0/1 prediction value calculation unit 1003 calculates a prediction value of either one of prediction modes 0, 1 of a 4×4 pixel block. A mode 3/4 prediction value calculation unit 1006 calculates a prediction value of either one of prediction modes 3, 4 of the 4×4 pixel block. A mode 5/6 prediction value calculation unit 1009 calculates a prediction value of either one of prediction modes 5, 6 of the 4×4 pixel block. A mode 7/8 prediction value calculation unit 1012 calculates a prediction value of either one of prediction modes 7, 8 of the 4×4 pixel block. At first, a prediction error is calculated based on a prediction mode due to prediction mode 0, 3, 5, 7; and in the next cycle, the prediction error is calculated based on a prediction mode due to the prediction modes 1, 4, 6, 8. Then, an intra prediction control unit 1019 decides the most efficient prediction mode according to prediction errors of the all prediction modes. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an image coding technique for moving images, and more particularly to intra-frame predictive coding.

  Conventionally, the JPEG system standardized by ISO (International Organization for Standardization) has been widely used as an encoding system for storing and displaying still images on a storage medium such as an optical disk such as a CD-ROM or a hard disk. In addition, the MPEG system standardized by ISO is widely used as an encoding system for storing and displaying moving images and audio on a similar storage medium, broadcasting over a communication channel, or bidirectional communication. Yes.

  The MPEG method, which is a moving image and audio encoding method, has been improved year by year for the purpose of improving the encoding efficiency. MPEG-1 in 1992, MPEG-2 in 1994, MPEG-2 in 1999 MPEG-4, and in 2003 MPEG-4AVC / H. H.264 is defined as an international standard.

  In MPEG-4, not only improvement in coding efficiency but also enhancement of a code error correction function and introduction of object coding for performing optimum coding for each coding target are realized. Furthermore, the system layer that defines the file format including the visual layer and the audio layer on the MPEG-4 syntax has a mechanism that allows easy access to moving image or audio data at any time requested by the user. providing.

  In addition, MPEG-4 AVC / H. H.264 employs a wide variety of new encoding tools for inter-frame prediction, intra-frame prediction, entropy encoding, and the like with the aim of further improving the encoding efficiency with respect to MPEG-4. In particular, in the intra-frame prediction focused on in the present invention, a prediction method in units of 4 × 4 image blocks and a prediction method in units of 16 × 16 image blocks that are effective for gradation images and the like as intra-frame prediction for luminance signals. Are defined for the 4 × 4 luminance block and four prediction modes for the 16 × 16 luminance block.

  FIG. 10 shows the relative positional relationship between the 4 × 4 luminance block and the reference image sample used for intra-frame prediction. The 4 × 4 luminance blocks to be encoded are (0,0), (1,0), (2,0), (3,0), (0,1), (1,1), ( 2,1), (3,1), (0,2), (1,2), (2,2), (3,2), (0,3), (1,3), (2, 3) and 16 samples of (3, 3). On the other hand, the reference image sample is located around the 4 × 4 luminance block and has been encoded. In the figure, (-1, -1), (0, -1), (1, -1), (2, -1), (3, -1), (4, -1), (5,- 1), (6, -1), (7, -1), (-1, 0), (-1, 1), (-1, 2), (-1, 3). ing.

  FIGS. 11 to 18 are conceptual diagrams showing the positional relationship between 16 samples to be encoded and reference samples used to calculate a prediction value for each sample in each prediction mode. Looking at the direction from the reference sample to be used to the encoding target image sample, mode 0 in FIG. 11 is predicted in the vertical direction, and mode 1 in FIG. 12 is predicted in the horizontal direction. Although not shown in the figure, mode 2 uses an average value of 13 reference samples as a predicted value. Mode 3 in FIG. 13 is prediction in the left diagonal downward direction, and mode 4 in FIG. 14 is prediction in the right diagonal downward direction. Mode 5 in FIG. 15 is prediction to the right vertical diagonally downward direction, and mode 6 in FIG. 16 is prediction to the right horizontal diagonally downward direction. Mode 7 in FIG. 17 is prediction in the left vertical diagonal downward direction, and mode 8 in FIG. 18 is prediction in the right horizontal diagonal upward direction.

Also, four prediction modes for 8 × 8 color difference blocks are prepared for intra-frame prediction for color difference signals. It is up to the user to decide which prediction mode to use for intra-frame prediction coding for either luminance signals or chrominance signals, and select the optimal prediction mode to achieve high coding efficiency. It is important to do.
ISO / IEC 14496-10; "Information technology--Coding of audio-visual objects--Part 10: Advanced video coding"; ISO / IEC; 2003-12-01

However, in order to determine which prediction mode to use among 13 prediction modes for luminance signals and 4 prediction modes for color difference signals, calculation of a prediction value used in the processing target block and prediction value from the processing target block are performed. It is necessary to calculate a prediction residual obtained by subtracting, and to calculate an index value such as SAD (Sum of Absolute Difference), and enormous operations are required to perform this for all prediction modes.
Furthermore, in order to perform intra-frame prediction of a 4 × 4 luminance block, the immediately preceding adjacent 4 × 4 luminance block is intra-frame predicted, orthogonally transformed, quantized, and then reconstructed by inverse quantization and inverse orthogonal transformation. In order to improve throughput processing performance, it is necessary to complete this series of processing in a short period of time.

  The present invention has been made in view of such problems, and an object of the present invention is to provide an intra-frame predictive coding technique that reduces computational resources necessary for prediction mode selection while realizing high throughput processing performance.

In order to solve this problem, the image coding apparatus of the present invention has the following configuration. That is,
An image in which a prediction error is calculated using one of a plurality of prediction modes for each pixel block composed of a plurality of pixels from one frame of input image data, and the calculated prediction error is encoded by an encoding unit. An encoding device comprising:
Among two or more prediction modes, one of the two or more prediction modes is used for two or more prediction modes having the same calculation content for calculating a prediction value and different pixel positions referred to for calculation of the prediction value. A plurality of prediction value calculation means for calculating a prediction value in the prediction mode,
A plurality of prediction error calculation means for calculating a prediction error of each pixel in the pixel block of interest based on the prediction values calculated by the plurality of prediction value calculation means;
A plurality of summing means for calculating the sum of absolute values of prediction errors of the respective pixels in the target pixel block calculated by the plurality of prediction error calculating means;
For each of the plurality of predicted value calculation means, a prediction mode that can be executed is sequentially set, and a predicted value is calculated for each of them, and the prediction error calculation means and the summation means perform summation. Prediction mode setting means for
Determining means for determining, as the prediction mode of the pixel block of interest, a prediction mode in which the minimum sum is calculated when the calculation of the sum of the absolute values of all prediction modes to be processed by the summing means is completed; .

  According to the present invention, it is possible to perform intraframe predictive coding while reducing computational resources necessary for prediction mode selection while realizing high throughput processing performance.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the embodiment, MPEG-4 AVC / H. An example applied to the H.264 encoding method will be described.

[First Embodiment]
FIG. 3 is a block configuration diagram of an image encoding device that performs encoding in units of pixel blocks in the first embodiment. This apparatus includes an intra-frame prediction unit 101, a prediction scheme selection unit 102, an orthogonal transformation unit 103, a quantization unit 104, and an entropy encoding unit 105. In addition, this apparatus includes an entropy decoding unit 106, an inverse quantization unit 107, an inverse orthogonal transform unit 108, a loop filter 109, a motion compensation unit 110, and an inter-frame decoding unit that decodes encoded data generated by the entropy encoding unit 105. A motion prediction unit 111 and a reference frame storage unit 112 are included. The input source of the moving image data may be a video camera or a storage medium storing moving image data.

  FIG. 1 is a block configuration diagram of the intra-frame prediction unit 101. The intra-frame prediction unit 101 includes a reference image data storage unit 1001 that stores a reference image in a block that has been encoded and adjacent to the target pixel block when encoding the target pixel block in one frame. . In addition, the apparatus includes a 4 × 4 block storage unit 1002 that stores pixel data (luminance data) in the block of interest. Other configurations are as described below.

  The mode 0/1 predicted value calculation unit 1003 calculates a predicted value of each pixel in the block of interest in the prediction mode 0 or 1. The mode 0/1 prediction error calculation unit 1004 calculates the prediction error of each pixel in the block of interest in the prediction mode 0 or 1. The mode 0/1 SAD calculation unit 1005 calculates a SAD (Sum of Absolute Difference) of the prediction error in the block of interest in the prediction mode 0 or 1.

  A mode 3/4 predicted value calculation unit 1006 calculates a predicted value in the prediction mode 3 or 4. The mode 3/4 prediction error calculation unit 1007 calculates the prediction error in the prediction mode 3 or 4. The mode 3/4 SAD calculation unit 1008 calculates the SAD of the prediction error in the block of interest in the prediction mode 3 or 4.

  A mode 5/6 predicted value calculation unit 1009 calculates a predicted value in the prediction mode 5 or 6. The mode 5/6 prediction error calculation unit 1010 calculates the prediction error in the prediction mode 5 or 6. The mode 5/6 SAD calculation unit 1011 calculates the SAD in the prediction mode 5 or 6.

  The mode 7/8 predicted value calculation unit 1012 calculates the predicted value in the prediction mode 7 or 8. A mode 7/8 prediction error calculation unit 1013 calculates a prediction error in prediction mode 7 or 8. The mode 7/8 SAD calculation unit 1014 calculates the SAD in the prediction mode 7 or 8.

  The mode 2 predicted value calculation unit 1015 calculates a predicted value in the prediction mode 2. The mode 2 prediction error calculation unit 1016 calculates a prediction error in the prediction mode 2. The mode 2 SAD calculation unit 1017 calculates the SAD in the prediction mode 2.

  The SAD storage unit 1018 temporarily stores the SAD calculated by each SAD calculation unit. The intra prediction control unit 1019 searches for the smallest SAD from each SAD stored in the SAD storage unit 1018 and determines the prediction mode in which the minimum SAD is calculated.

  Usually, the smaller the prediction error is, the higher the coding efficiency is by assigning a codeword having a shorter code length. In other words, it can be said that the smaller the SAD, the higher the encoding efficiency. Therefore, the intra prediction control unit 1019 searches for the smallest SAD, determines which of the prediction modes having the minimum value is 0 to 8, and sets the determination result as the final prediction mode.

  The selection unit 1020 has an internal buffer for storing the prediction error from each prediction error calculation unit, selects a prediction error corresponding to the prediction mode determined by the intra prediction control unit 1019, and sends it to the prediction method selection unit 1020. Output. At this time, the selection unit 1020 also outputs information specifying the selected prediction mode to the prediction method selection unit 102.

  As described above, prediction modes 0 and 1, modes 3 and 4, modes 5 and 6, and modes 7 and 8 are performed in a common configuration, although pixel positions to be referred to for prediction value calculation are different. This is because the calculation contents for calculating the predicted value are the same. Here, MPEG-4AVC / H. Since H.264 is taken as an example, the maximum number of prediction modes that can be realized with one configuration is two. However, in the case of an encoding method having two or more prediction modes with common operation contents, naturally one configuration is used. It is possible to support two or more prediction modes.

  FIG. 4 is a flowchart showing an operation control sequence of the entire intra-frame prediction unit 101 by the intra prediction unit 1019. FIG. 6 illustrates intra prediction → integer conversion → quantization → inverse quantum by the intra-frame prediction unit 101 of FIG. 3 including the reference image data storage unit 1001 illustrated in FIG. 1 or the prediction error selection unit 1020. 4 is a timing chart showing the number of processing cycles required for each process in a series of operations from conversion to inverse integer conversion in time series.

  Hereinafter, with regard to the operation mode in the embodiment of the image encoding device, a series of operations related to the encoding process, particularly a prediction encoding process procedure until a prediction mode is determined by the intra-frame prediction unit 101 and a prediction error is output. This will be described with reference to FIGS. 1, 3, 4, and 6.

In the intra-frame prediction unit 101 of FIG. 3 configured by the reference image data storage unit 1001 to the difference selection unit 1020 shown in FIG. 1, as shown in FIG. 6, intra prediction → integer conversion → quantization → inverse quantization → In processing cycle 0 (see FIG. 6), a series of operations up to inverse integer conversion is processed in a total of 9 processing cycles. Intra prediction control unit 1019 sets prediction mode 0 to the mode 0/1 predicted value calculation unit, Prediction mode 3 for mode 3/4 prediction value calculation unit 1006, prediction mode 5 for mode 5/6 prediction value calculation unit 1009, and prediction mode 7 for mode 7/8 prediction value calculation unit 1012 Is set (step S101). That is, the intra prediction control unit 1019 functions as a prediction mode setting unit for each prediction value calculation unit.

  The intra prediction control unit 1019 also notifies the SAD storage unit 1018 that the prediction mode 0/3/5/7 has been selected. Upon receiving this notification, the SAD storage unit 1018 stores the SAD that is the summation result at the address position defined in advance for each prediction mode.

  The mode 0/1 predicted value calculation unit 1003 calculates a predicted value in the prediction mode 0 using the reference sample input from the reference image data storage unit 1001. The mode 0/1 prediction error calculation unit 1004 calculates the prediction error using the block image data whose prediction value is input from the 4 × 4 block storage unit 1002. The calculated prediction errors (4 × 4) are output to the mode 0/1 SAD calculation unit 1005 and the selection unit 1020. The selection unit 1020 temporarily stores this prediction error in an internal buffer. The mode 0/1 SAD calculation unit 1005 calculates the SAD using the input prediction error and outputs it to the SAD storage unit 1018. Similarly, SADs calculated in the prediction modes 3, 5, and 7 are stored in the SAD storage unit 1018, and prediction errors in the prediction modes 3, 5, and 7 are stored in the internal buffer of the selection unit 1020.

  Further, the prediction mode 2 prediction value calculation unit 1015, the prediction mode 2 prediction error calculation unit 1016, and the prediction mode 2 SAD calculation unit 1017 execute the processing in the processing cycle 0 (which may be processing cycle 1 described below). Shall. Accordingly, in cycle 0, the SADs of the prediction modes 0, 2, 3, 5, and 7 are stored in the SAD storage unit 1018, and the prediction errors are stored in the internal buffer of the selection unit 1020, respectively.

  In the processing cycle 1 (see FIG. 6), the intra prediction control unit 1019 sets the prediction mode 1 for the mode 0/1 predicted value calculation unit 1003. Also, the prediction mode 4 for the mode 3/4 prediction value calculation unit 1006, the prediction mode 6 for the mode 5/6 prediction value calculation unit 1009, and the mode 7/8 prediction value calculation unit 1012. Sets the prediction mode 8 (step S102).

  The intra prediction control unit 1019 also notifies the SAD storage unit 1018 that the prediction mode 1/4/6/8 has been selected. A mode 0/1 prediction value calculation unit 1003 calculates a prediction value in prediction mode 1 using a reference sample input from the reference image data storage unit 1001, and a mode 0/1 prediction error calculation unit 1004 converts the prediction value to 4 ×. A prediction error is calculated using block image data input from the 4-block storage unit 1002. The calculated prediction error is output to the selection unit 1020 and the mode 0/1 SAD calculation unit 1004. The mode 0/1 SAD calculation unit 1005 calculates the SAD using the input prediction error and outputs it to the SAD storage unit 1018. Similarly, SADs calculated in the prediction modes 4, 6, 8, and 2 are stored in the SAD storage unit 1018, and the prediction error of each prediction mode is stored in the internal buffer of the selection unit 1020.

  As described above, the calculation of SAD in all prediction modes ends.

  Thereafter, in the processing cycle 2 (see FIG. 6), the intra prediction control unit 1019 responds to the SAD calculated by all the prediction modes 0 to 8 stored in the SAD storage unit 1018 by the processing cycles 0 and 1. The prediction mode with the smallest SAD value is determined as the final prediction mode (step S103). Next, the intra prediction control unit 1019 notifies the selected prediction mode to the selection unit 1020 (step S104). As a result, the selection unit 1020 outputs a prediction error corresponding to the determined prediction mode among prediction errors of each prediction mode stored in the internal buffer, and also outputs information indicating the prediction mode (step S105). . When this output process is completed, the selection unit 1020 sets the prediction error of each prediction mode stored in the internal buffer so that it can be discarded or overwritten.

Next, the mode 0/1 predicted value calculation unit 1003, the mode 0/1 prediction error calculation unit 1004, and the mode 0/1 SAD calculation unit 1005 illustrated in FIG. 1 will be described with reference to FIG.

In the figure, 3001 is a reference sample (0, -1) stored in the reference image data storage unit 1001, and 3002 is a reference sample (-1, 0) (see FIG. 10). 3003 is an image sample (0,0) stored in the 4 × 4 block storage unit 1002, 3004 is an image sample (0,1), 3005 is an image sample (1,0), and 3006 is an image sample (0,2). , 3007 are image samples (2, 0), 3008 are image samples (0, 3), and 3009 are image samples (3, 0).

  Note that the coordinates of these samples are relative to the reference position in the pixel block of interest. For example, in the prediction mode 0, the reference position in the pixel block of interest is (i, 0) (i = 0, 1, 2, 3). Accordingly, for example, the pixel data (luminance data) indicated by the sample (0, −1) is the pixel value at the coordinates (0 + i, −1 + 0), that is, the coordinates (i, −1). Similarly, the sample (0, 3) has a pixel value of coordinates (i, 3).

  In mode 0, i = 0, 1, 2, 3 changes in four stages, so when i = 0, four prediction errors are calculated by the dividers 3010 to 3013, and their absolute values are calculated by ABS calculation. Calculation is performed by the units 3014 to 3017. Further, the absolute values are added by adders 3018 to 3020. The adder 3021 adds the values held in the register 3022 and stores the addition result in the register 3022 again. This register 3022 is initialized to “0” when calculating the SAD of one block.

  When the above processing is executed for i = 0, 1, 2, 3 and i = 3, the adder 3021 calculates the SAD of the block of interest. Therefore, when i = 3, the addition result of the adder 3021 is stored in the SAD storage unit 1018 as the mode 0 SAD of the block of interest.

  On the other hand, in mode 1, the reference position of the block of interest is (0, i) (i = 0, 1, 2, 3). Note that the changing coordinates are on the Y-axis (vertical axis). Except for this point, the processing of each processing unit is the same as mode 0.

  As described above, in prediction mode 0/1, it is not necessary to perform an operation to obtain a predicted value, and a desired reference sample is used as it is. The dividers 3010 to 3013 correspond to the mode 0/1 prediction error calculation unit 1004, and the ABS calculators 3014 to 1017 and the adders 3018 to 3019 that convert to absolute values correspond to the mode 0/1 SAD calculation unit 1005. Note that the prediction mode selection circuit, the prediction error calculation circuit, and the SAD calculation circuit are shared in prediction modes 0 and 1.

  FIG. 9 is a block diagram showing details of the mode 3/4 predicted value calculation unit 1006, the mode 3/4 prediction error calculation unit 1007, and the mode 3/4 SAD calculation unit 1008 shown in FIG.

  In the figure, reference numerals 4001 to 4006 denote reference samples stored in the reference image data storage unit 1001 (see FIG. 10). 4007 is a constant 2. Reference numerals 4008 to 4015 denote image samples stored in the 4 × 4 block storage unit 1002 (see FIG. 10).

  4016 to 4020 which are circuits for calculating the prediction value of the prediction mode 3/4 correspond to the mode 3/4 prediction value calculation unit 1006. The dividers 4021 to 4024 correspond to the mode 3/4 prediction error calculation unit 1007. Also, the ABS calculators 4025 to 4028, the adders 4029 to 4032, and the register 4033 that convert to absolute values correspond to the mode 3/4 SAD calculation unit 1007. The shifter 4017 is a circuit that shifts the input value by one bit in the upper direction, that is, doubles the input value. Similarly, the shifter 4020 is a circuit that shifts the input value by 2 bits in the lower direction, that is, makes the input value ¼.

  The operation in this configuration is the same as in the prediction mode 0/1 described above, and the prediction mode selection circuit, the prediction error calculation circuit, and the SAD calculation circuit are shared in the prediction modes 3 and 4. Note that The addition of the constant “2” by the adder 4018 is for rounding up the fractional part when the shifter 4020 makes the quarter.

  From the above description, it will be apparent that the other prediction modes 5 to 8 can be similarly configured.

  In the above-described embodiment, the intra-frame prediction encoding process for the 4 × 4 luminance block has been described as the intra-frame prediction unit 101. However, the intra-frame prediction encoding process for the 16 × 16 luminance block and the 8 × 8 color difference block is also described. The same applies.

  As described above, according to the present embodiment, the prediction mode is sequentially set in each prediction value calculation unit, and the prediction error of all prediction modes is calculated. As a result, the intra prediction encoding process can be realized on a scale substantially ½ compared to the case where the circuit configuration is independent for each mode.

[Second Embodiment]
Next, MPEG-4 AVC / H. A second embodiment applied to H.264 encoding will be described.
The overall configuration of the apparatus is the same as that in FIG. 3, and the intra-frame prediction unit 101 will be described below.

  FIG. 2 is a block configuration diagram of the intra-frame prediction unit 101 of the image coding apparatus to which the second embodiment is applied.

  The apparatus includes a reference image data storage unit 2001, a 4 × 4 block storage unit 2002, and the following configuration.

  The mode 0/1 predicted value calculation unit 2003 calculates a predicted value in the prediction mode 0 or 1. A mode 0/1 prediction error calculation unit 2004 calculates a prediction error in the prediction mode 0 or 1. A mode 01/2 SAD calculation unit 2005 calculates a SAD (Sum of Absolute Difference) in the prediction mode 0 or 1.

  The mode 3/4 predicted value calculation unit 2006 calculates the predicted value in the prediction mode 3 or 4. A mode 3/4 prediction error calculation unit 2007 calculates a prediction error in the prediction mode 3 or 4. The SAD calculation unit 2008 calculates a SAD (Sum of Absolute Difference) in the prediction mode 3 or 4.

  The mode 7/8 predicted value calculation unit 2009 calculates the predicted value in the prediction mode 7 or 8. The mode 7/8 prediction error calculation unit 2010 calculates the prediction error in the prediction mode 7 or 8. The mode 7/8 SAD calculation unit 2011 calculates the SAD in the prediction mode 7 or 8.

  The mode 5 predicted value calculation unit 2012 calculates the predicted value in the prediction mode 5. A mode 5 prediction error calculation unit 2013 calculates a prediction error in the prediction mode 5. The mode 5 SAD calculation unit 2014 calculates the SAD in the prediction mode 5.

  The mode 6 predicted value calculation unit 2015 calculates a predicted value in the prediction mode 6. A mode 6 prediction error calculation unit 2016 calculates a prediction error in the prediction mode 6. The mode 6 SAD calculation unit 2017 calculates the SAD in the prediction mode 6.

  The mode 2 predicted value calculation unit calculates a predicted value in the prediction mode 2. A mode 2 prediction error calculation unit 2019 calculates a prediction error in the prediction mode 2. The mode 2 SAD calculation unit 2020 calculates the SAD in the prediction mode 2.

  The SAD storage unit 2021 temporarily stores the SAD calculated by each SAD calculation unit. Furthermore, the intra prediction control unit 2022 searches for the smallest SAD from among the SADs stored in the SAD storage unit 2021, and determines the prediction mode. The selection unit 2023 has a buffer memory therein in order to store the prediction error calculated by each prediction error calculation unit. Then, the selection unit 2023 reads the prediction error corresponding to the prediction mode determined by the intra prediction control unit 2022 from the buffer memory and outputs it. At this time, information specifying the determined prediction mode is also output. When the output of the prediction error for one block is completed, the selection unit 2023 enables the internal buffer to be erased or overwritten.

  FIG. 5 is a flowchart showing an operation control sequence of the entire intra-frame prediction unit 101 by the intra prediction control unit 2022. 7 shows intra prediction → integer conversion → quantization → inverse quantization → inverse integer by the intra-frame prediction unit 101 in FIG. 3 including the reference image data storage unit 2001 to the prediction error selection unit 2023 shown in FIG. 5 is a timing chart showing the number of processing cycles required for each process in a series of operations up to conversion in time series.

  Hereinafter, with regard to the operation mode in the embodiment of the image encoding device, a series of operations related to the encoding process, particularly a prediction encoding process procedure until a prediction mode is determined by the intra-frame prediction unit 101 and a prediction error is output. This will be described with reference to FIGS. 2, 3, 5, and 7.

  In the intra-frame prediction unit 101 of FIG. 3 configured by the reference image data storage unit 2001 to the selection unit 2023 shown in FIG. 2, intra prediction → integer conversion → quantization → inverse quantization → inverse as shown in FIG. A series of operations up to integer conversion is processed in a total of 9 processing cycles, but unlike FIG. 6, the next 4 × 4 block intra-frame predictive encoding process can be started when processing cycle 6 is completed. It is. In processing cycle 7, it is noted that the inverse integer conversion process of 4 × 4 luminance block # 1 and the intra-frame prediction process for prediction mode 0/3/7 of 4 × 4 luminance block # 2 are executed simultaneously. I want. Note that this is possible because prediction mode 0/3/7 does not use the reconstructed image for 4 × 4 luminance block # 1 as a reference sample.

  In the processing cycle “−1” (see FIG. 7), the intra prediction control unit 2022 performs the prediction mode 0 for the mode 0/1 predicted value calculation unit 2003 and the mode 3/4 predicted value calculation unit 2006. The prediction mode 3 is preferentially set for the mode 7/8 predicted value calculation unit 2009 (step S201). The intra prediction control unit 2022 also notifies the SAD storage unit 2021 that the prediction mode 0/3/7 has been selected. The mode 0/1 predicted value calculation unit 2003 calculates a predicted value in the prediction mode 0 using the reference sample input from the reference image data storage unit 2001. A mode 0/1 prediction error calculation unit 2004 calculates a prediction error using block image data whose prediction value is input from the 4 × 4 block storage unit 2002. Further, the mode 0/1 SAD calculation unit 2005 calculates SAD using the prediction error, and outputs the SAD to the SAD storage unit 2021 and the selection unit 2023. Hereinafter, the SAD calculated in the prediction modes 3 and 7 is similarly stored in the SAD storage unit 2021.

  In the processing cycle 0 (see FIG. 7), the intra prediction control unit 2022 uses the prediction mode 1 for the mode 0/1 prediction value calculation unit 2003 and the prediction mode 4 for the mode 3/4 prediction value calculation unit 2006. For the mode 7/8 predicted value calculation unit 2009, the prediction mode 8 is selected (S202). The intra prediction control unit 2022 also notifies the SAD storage unit 2021 that the prediction mode 1/4/8 has been selected. A mode 0/1 prediction value calculation unit 2003 calculates a prediction value in prediction mode 1 using a reference sample input from the reference image data storage unit 2001, and a mode 0/1 prediction error calculation unit 2004 uses the prediction value as 4 ×. A prediction error is calculated using block image data input from the 4-block storage unit 2002. Further, the mode 0/1 SAD calculation unit 2005 calculates SAD using the prediction error, and outputs the SAD to the SAD storage unit 2021. In the same manner, SADs calculated in the prediction modes 4, 8, 5, 6, and 2 are similarly stored in the SAD storage unit 2021.

  In the processing cycle 1 (see FIG. 7), the intra prediction control unit 2022 minimizes the SAD calculated by all the prediction modes 0 to 8 stored in the SAD storage unit 2021 by the processing cycles −1 and 0. The SAD value is searched, and the prediction mode in which the reference SAD value is calculated is determined as the final prediction mode (step S203). Next, the intra prediction control unit 2022 notifies the selection unit 2023 of the selected prediction mode (step S204). As a result, the selection unit 2023 outputs the prediction error of the corresponding prediction mode and information specifying the prediction mode stored in the internal buffer memory (step S205).

  As described above, the second embodiment is configured to achieve a higher throughput than the first embodiment.

  In the second embodiment, the intra-frame prediction encoding process for the 4 × 4 luminance block is described as the intra-frame prediction unit 101. However, the intra-frame prediction code for the 16 × 16 luminance block and the 8 × 8 color difference block is described. The present invention can also be applied in the same way.

[Other Embodiments]
In the first and second embodiments, the example realized by hardware has been described, but may be realized by a computer program executed by a computer. In this case, since the computer program can share the same procedure (or function) in a plurality of modes, the size of the computer program can be small.

  In addition, the computer program is usually stored in a computer-readable storage medium such as a CD-ROM, and is set in a reading unit (CD-ROM drive or the like) of the computer and copied or installed in the system. It becomes possible. Therefore, it is obvious that such a computer readable storage medium falls within the scope of the present invention.

It is a block block diagram of the intra-frame prediction part in 1st Embodiment. It is a block block diagram of the intra-frame prediction part in 2nd Embodiment. It is a block block diagram of the image coding apparatus in embodiment. It is a flowchart which shows the process sequence of the estimation part in a flame | frame in 1st Embodiment. It is a flowchart which shows the process sequence of the prediction part in a flame | frame in 2nd Embodiment. It is a figure which shows the sequence of the encoding process in 1st Embodiment. It is a figure which shows the sequence of the encoding process in 2nd Embodiment. It is a block block diagram which shows the detail of the mode 0/1 prediction value calculation part in 1st Embodiment, the mode 0/1 prediction error calculation part, and the mode 0/1 SAD calculation part. It is a block block diagram which shows the detail of the mode 3/4 prediction value calculation part in 1st Embodiment, the mode 3/4 prediction error calculation part, and the mode 3/4 SAD calculation part. It is a figure showing the relative positional relationship of the pixel in a 4x4 brightness | luminance block, and the reference image sample used for prediction. It is a figure which shows the prediction direction in the prediction mode 0. FIG. It is a figure which shows the prediction direction in the prediction mode 1. FIG. It is a figure which shows the prediction direction in the prediction mode 3. FIG. It is a figure which shows the prediction direction in the prediction mode 4. FIG. It is a figure which shows the prediction direction in the prediction mode 5. FIG. It is a figure which shows the prediction direction in the prediction mode 6. FIG. It is a figure which shows the prediction direction in the prediction mode 7. FIG. It is a figure which shows the prediction direction in the prediction mode 8. FIG.

Claims (4)

An image in which a prediction error is calculated using one of a plurality of prediction modes for each pixel block composed of a plurality of pixels from one frame of input image data, and the calculated prediction error is encoded by an encoding unit. An encoding device comprising:
Among two or more prediction modes, one of the two or more prediction modes is used for two or more prediction modes having the same calculation content for calculating a prediction value and different pixel positions referred to for calculation of the prediction value. A plurality of prediction value calculation means for calculating a prediction value in the prediction mode,
A plurality of prediction error calculation means for calculating a prediction error of each pixel in the pixel block of interest based on the prediction values calculated by the plurality of prediction value calculation means;
A plurality of summing means for calculating the sum of absolute values of prediction errors of the respective pixels in the target pixel block calculated by the plurality of prediction error calculating means;
For each of the plurality of predicted value calculation means, a prediction mode that can be executed is sequentially set, and a predicted value is calculated for each of them, and the prediction error calculation means and the summation means perform summation. Prediction mode setting means for
Determining means for determining, as the prediction mode of the pixel block of interest, a prediction mode in which the minimum sum is calculated when the summation of the absolute values of all prediction modes to be processed by the summing means is completed. An image encoding apparatus characterized by that. Storage means for storing prediction errors calculated by the plurality of prediction error calculation means;
The image according to claim 1, further comprising: a selecting unit that selects a prediction error corresponding to the prediction mode determined by the determining unit from among those stored in the storage unit, and outputs the selected error to the encoding unit. Encoding device. Furthermore, in order to obtain a pixel value of a pixel to be referred to in each prediction mode, a decoding unit that decodes the encoded data obtained by the encoding unit is provided.
A prediction mode for obtaining a prediction value by referring to a pixel in a pixel block immediately before the pixel block of interest and a prediction mode for obtaining a prediction value by referring to a pixel in a block encoded before the previous pixel block The prediction mode setting means preferentially sets a prediction mode for obtaining a prediction value by referring to a pixel in a block encoded before the previous pixel block. The image coding apparatus according to claim 1 or 2. Among two or more prediction modes, calculation contents for calculating a prediction value are common, and two or more prediction modes having different pixel positions to be referred to for prediction value calculation are either one of the two or more prediction modes. A plurality of prediction value calculation means for calculating a prediction value in the prediction mode;
A plurality of prediction error calculation means for calculating a prediction error of each pixel in a pixel block composed of a plurality of pixels based on the prediction values calculated by the plurality of prediction value calculation means;
A plurality of summing means for calculating the sum of absolute values of prediction errors of the respective pixels in the pixel block calculated by the plurality of prediction error calculating means;
A method for controlling an image coding apparatus, wherein a prediction error is calculated from one frame of input image data using any one of a plurality of prediction modes for each pixel block, and the calculated prediction error is encoded by an encoding unit. Because
For each of the plurality of predicted value calculation means, a prediction mode that can be executed is sequentially set, and a predicted value is calculated for each of them, and the prediction error calculation means and the summation means perform summation. Prediction mode setting process
A determination step of determining, as the prediction mode of the pixel block of interest, a prediction mode in which a minimum summation result is calculated when the calculation of the sum of the absolute values of all prediction modes to be processed by the summing unit is completed. A control method for an image encoding device.

Images