JP5219062B2 - Image data generation method - Google Patents

Description

  The present invention relates to a method for generating image data, and more particularly to a method for generating image data including intra prediction for easily determining a prediction mode.

  The latest international standard for video coding by VCEG (Video Coding Experts Group) of ITU-T (International Telecommunication Union-Telecommunication Standardization Sector) and ISO / IEC Moving Picture Expetrts Group (MPEG) As H. H.264 has been developed and put into practical use.

  H. H.264 can compress twice as much as the same image quality as MPEG-2 and MPEG-4, and can be applied to a wide range of applications from low bit rate video conferencing to HDTV (High Definition Television). is there.

  H. One of the characteristics of H.264 is a method called intra-frame prediction (intra prediction). This is a method for improving the compression rate using pixel correlation, and is a method for generating a predicted image by interpolating data between sub-blocks. By referring to the block, prediction is performed by comparison at the pixel level.

  Intra prediction is also employed in MPEG-4, but MPEG-4 refers to the left, upper, and upper left subblocks of the encoding target subblock. In the case of H.264, since the left, upper, upper left, and upper right sub-blocks are referred to, a predicted image with higher accuracy can be obtained.

  In MPEG-4, prediction is performed in units of 8 × 8 pixels. In H.264, complicated images are predicted in units of 4 × 4 pixels, and simple images are predicted in units of 16 × 16 pixels. Therefore, efficient prediction is possible.

  H. In intra prediction in H.264, the optimal prediction mode is determined by using four prediction methods (prediction modes) for prediction in 16 × 16 pixel units, and nine predictions in prediction in 4 × 4 pixel units. Since the optimum prediction mode is determined using the mode, there is a problem that enormous calculation processing is required and it takes time until the prediction result is obtained.

  For example, Patent Literature 1 discloses a technique for achieving intra prediction time reduction by simultaneously performing intra prediction on a plurality of encoding target sub-blocks.

JP 2005-130509 A

  As described above, it is necessary to reduce the time spent for intra prediction as much as possible. H.264, and one of the solutions has been proposed in Japanese Patent Application Laid-Open No. H11-260, but the overall time is shortened by providing a plurality of calculators for performing intra prediction. There was nothing to reduce the time spent in intra prediction of sub-blocks.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a method for generating image data in which the intra prediction of each sub-block is made efficient and the time spent for intra prediction is shortened. And

According to a first aspect of the present invention, there is provided an image data generation method including an intra prediction for predicting an image of a second sub-block to be encoded using pixel values of a plurality of encoded first sub-blocks. A method of generating image data having prediction, wherein the pixel before encoding of the first sub-block when encoding of the first sub-block is not completed prior to the intra prediction Based on the result of the interpolation for each of the plurality of prediction modes, the value is a first prediction value, interpolation is performed for each of the plurality of prediction modes for the second sub-block using the first prediction value. A simple intra prediction that selects at least one of the plurality of prediction modes as an optimal prediction mode candidate, and the intra prediction performs the image prediction using the at least one optimal prediction mode candidate. The intra prediction is performed after encoding is completed for all of the first sub-blocks, and the pixel values after encoding of the first sub-blocks for which encoding is completed are used as second prediction values, Performing interpolation for each of the at least one optimum prediction mode candidate for the second sub-block using the second prediction value, and based on the result of interpolation for each of the at least one optimum prediction mode candidate, There line the image prediction to determine the optimum prediction mode from the at least one optimum prediction mode candidate, the in simplified intra prediction, for the first sub-block coding is completed, the first The pixel value after encoding of the sub-block or the pixel value before encoding is used as the first predicted value .

According to a second aspect of the present invention, in the simple intra prediction, the first predicted image of the second sub-block obtained as a result of interpolation for each of the plurality of prediction modes; By acquiring the absolute value sum of the difference from the pixel value before the interpolation of the second sub-block is performed for each of the plurality of prediction modes, and selecting in order from the smallest absolute value sum, At least one optimum prediction mode candidate is selected.

According to a third aspect of the present invention, there is provided the method for generating image data according to the second prediction image of the second sub-block obtained as a result of interpolation for each of the at least one optimum prediction mode candidate. The absolute value sum of the difference from the pixel value before the interpolation of the second sub-block is executed is obtained for each of the at least one optimum prediction mode candidate, and the one having the smallest absolute value sum is selected. Thus, the optimum prediction mode is determined.

According to a fourth aspect of the present invention, in the image data generation method, the simple intra prediction and the intra prediction are executed by pipeline processing, and the pipeline processing is performed with respect to the nth subblock. In the n-th processing cycle in which the intra-prediction is performed, the intra-prediction is performed on the n-1-th sub-block for which the simple intra-prediction is performed in the n-1th processing cycle, and the intra Using the optimal prediction mode determined by prediction, local decoding is performed on the n-1st sub-block.

According to the image data generation method of the first aspect of the present invention, when the encoding of the first sub-block is not completed, the pixel value before the encoding of the first sub-block is set to the first Using the first prediction value as a prediction value, interpolation is performed for each of the plurality of prediction modes for the second sub-block, and based on the result of the interpolation for each of the plurality of prediction modes, from the plurality of prediction modes Since it has simple intra prediction that selects at least one as an optimal prediction mode candidate, it is possible to start image prediction even at a stage where coding has not been completed for all of the first sub-blocks that are reference target sub-blocks. Thus, it is possible to improve the efficiency of intra prediction of individual sub-blocks. In addition, the time required for intra prediction can be shortened , and even when there is a coding that has been completed and a coding that has not been completed in the first sub-block, the optimum prediction mode candidate is appropriately selected. it is Ru can.

According to the image data generation method of the second aspect of the present invention, the optimum prediction mode candidate can be appropriately selected in the simple intra prediction.

According to the image data generation method of the third aspect of the present invention, it is possible to appropriately determine the optimum prediction mode in intra prediction.

According to the image data generation method of the fourth aspect of the present invention, it is possible to efficiently perform the encoding process of the image data by executing the simple intra prediction and the intra prediction by pipeline processing.

<Intra prediction mode>
Prior to the description of the embodiments of the invention, H.C. In the intra prediction in H.264, nine prediction modes used for prediction in units of 4 × 4 pixels will be described.

  FIG. 1 is a diagram schematically showing a macro block MB composed of 16 × 16 pixels, and the macro block MB is composed of 16 sub-blocks SB.

  Each sub-block SB is composed of 4 × 4 pixels, and intra prediction and encoding are executed in units of sub-blocks. In FIG. 1, the numbers assigned to each sub-block SB indicate the encoding order, and the sub-block in the upper left corner is numbered 0, the right next to it is number 1 and the number below 0 is numbered zigzag. It has been.

  Further, a block BL of 8 × 8 pixel units is constituted by four sub blocks SB, and the 0th to 3rd sub blocks SB, the 4th to 7th sub blocks SB, and the 8th to 11th sub blocks SB The 12th to 15th sub-blocks SB constitute one block BL.

  Hereinafter, assuming the case where intra prediction is performed on the twelfth sub-block SB, nine prediction modes will be described with reference to FIGS.

  FIG. 2 is a diagram in which the symbols a to p are assigned to the 16 encoding target pixels UP constituting the 12th sub-block SB for convenience, and the pixel in the upper left corner is designated as a. The right-hand side is b, the c and d are assigned consecutively, the pixels below the a are the e, and the same rules are used up to p.

  In addition, among the encoded pixels EP in the encoded sub-block adjacent to the 12th sub-block SB, pixels used as reference pixels in the intra prediction are denoted by symbols A to M for convenience. Assigned.

  That is, A to D are assigned to the encoded pixels EP in the sixth sub-block SB (upper block) shown in FIG. 1 in order from the left, and the encoded pixels in the seventh sub-block SB (upper right block) are assigned. EP is assigned E to H in order from the left, and the encoded pixels EP in the ninth sub-block SB (left block) are assigned I to L in order from the top, and the third sub-block SB (upper left) M is assigned to the encoded pixel EP in the block).

  FIG. 3 is a diagram schematically showing a prediction method called the prediction mode 0. As indicated by arrows, the pixel values of the A to D encoded pixels EP are expressed as A to D, respectively. A prediction image is obtained by performing interpolation using the prediction value of the encoding target pixel UP existing in the vertical direction of each of the encoded pixels EP. Whether or not the predicted image is appropriate is determined by calculating a difference from the initial pixel value (pixel value before the interpolation is performed) of the encoding target pixel UP and by determining the magnitude of the absolute value sum thereof.

  FIG. 4 is a diagram schematically illustrating a prediction method referred to as prediction mode 1. As indicated by arrows, the pixel values of I to L encoded pixels EP are expressed as I to L, respectively. The prediction image is obtained by performing interpolation using the prediction value of the encoding target pixel UP existing in the horizontal direction of each of the encoded pixels EP. Whether or not the predicted image is appropriate is determined by calculating the difference from the original pixel value of the encoding target pixel UP and determining the magnitude of the sum of absolute values thereof.

  FIG. 5 is a diagram schematically illustrating a prediction method referred to as prediction mode 2, in which the average values of the respective pixel values of the A to D and I to L encoded pixels EP are represented by all codes. A prediction image is obtained by executing interpolation using the prediction value of the pixel to be converted UP, and is also referred to as a DC (direct current) mode. Whether or not the predicted image is appropriate is determined by calculating the difference from the original pixel value of the encoding target pixel UP and determining the magnitude of the sum of absolute values thereof.

  FIG. 6 is a diagram schematically showing a prediction method called prediction mode 3, in which the pixel values of B to H encoded pixels EP are expressed as B to H as indicated by arrows. A predicted image is obtained by performing interpolation using the encoded pixel EP as a predicted value of the encoding target pixel UP existing at a position in the diagonally downward left direction of each encoded pixel EP. Whether or not the predicted image is appropriate is determined by calculating the difference from the original pixel value of the encoding target pixel UP and determining the magnitude of the sum of absolute values thereof.

  FIG. 7 is a diagram schematically showing a prediction method called prediction mode 4, and each pixel value of encoded pixels EP of A to C, M, I, and J as indicated by arrows. Is used as the predicted value of the encoding target pixel UP existing at the position of the diagonally lower right direction of each of the encoded pixels EP of A to C, M, I, and J, and the prediction image is executed. obtain. Whether or not the predicted image is appropriate is determined by calculating the difference from the original pixel value of the encoding target pixel UP and determining the magnitude of the sum of absolute values thereof.

  FIG. 8 is a diagram schematically showing a prediction method called prediction mode 5, and as indicated by arrows, respective pixel values of M, J, and A to C encoded pixels EP are expressed as follows. A predicted image is obtained by performing interpolation using the predicted values of the encoding target pixel UP existing at positions on the right side of the vertical direction of each of the M, J, and A to C encoded pixels EP. Whether or not the predicted image is appropriate is determined by calculating the difference from the original pixel value of the encoding target pixel UP and determining the magnitude of the sum of absolute values thereof.

  FIG. 9 is a diagram schematically showing a prediction method called prediction mode 6, and as indicated by arrows, respective pixel values of M and I to K encoded pixels EP are represented by M and A prediction image is obtained by performing interpolation using the prediction values of the encoding target pixel UP existing at positions lower than the horizontal direction of each of the I to K encoded pixels EP. Whether or not the predicted image is appropriate is determined by calculating the difference from the original pixel value of the encoding target pixel UP and determining the magnitude of the sum of absolute values thereof.

  FIG. 10 is a diagram schematically showing a prediction method called prediction mode 7, and as indicated by arrows, the respective pixel values of the encoded pixels EP of B to E are expressed as B to E. The prediction image is obtained by performing interpolation using the prediction value of the encoding target pixel UP present at the position on the left side of each of the encoded pixels EP in the vertical direction. Note that the suitability of the predicted image is determined by calculating the difference between the original pixel value of the encoding target pixel UP and the magnitude of the sum of absolute values thereof.

  FIG. 11 is a diagram schematically showing a prediction method called the prediction mode 8, and as indicated by arrows, the pixel values of the encoded pixels EP of J to L are represented by J to L, respectively. The prediction image is obtained by performing interpolation using the prediction value of the encoding target pixel UP existing at a position higher than the horizontal direction of each of the encoded pixels EP. Whether or not the predicted image is appropriate is determined by calculating the difference from the original pixel value of the encoding target pixel UP and determining the magnitude of the sum of absolute values thereof.

  H. In intra prediction in H.264, prediction is performed using the nine prediction modes described above, and the prediction mode with the smallest sum of absolute values of the differences obtained in each mode is determined to be the optimal prediction mode. The encoding target pixel UP is encoded using this mode.

  In addition, as described with reference to FIGS. 3 to 11, in order to perform intra prediction, reference is made to the left, upper, upper left, and upper right subblocks of the encoding target subblock, and the encoding included in them Since the pixel value of the completed pixel EP is used as a predicted value, all of these sub-blocks need to be encoded.

  However, in encoding, as will be described later, it is necessary to sequentially perform processing such as transformation, quantization, inverse quantization, and inverse transformation, and all the encoding target pixels UP of one subblock SB are encoded. Requires a lot of processing time. This increases as the quality of moving images increases, and it is expected that the time will be further extended in HDTV.

  In addition, referring to FIG. 1, for example, when performing intra prediction on the 12th sub-block SB, the 3rd, 6th, 7th and 9th sub-blocks SB are changed as described above. Although referred to, these are encoded at the time when intra prediction of the 12th subblock SB is performed, and there is no problem in the intra prediction of the 12th subblock SB, but for example, for the 13th subblock SB When intra prediction is performed, all nine prediction modes cannot be executed unless encoding of the 12th sub-block SB is completed.

  For the 14th sub-block SB, unless the 13th sub-block SB has been encoded, the modes referring to the sub-block SB, that is, the prediction modes 3 and 7 cannot be executed.

  In the 0th and 1st sub-blocks SB, the situation differs depending on where the macroblock MB is located in the screen.

  That is, when the macro block MB constitutes the upper left part of the screen, there is no sub-block that can be referred to in the 0th sub-block SB. Make predictions.

  Further, in the first sub-block SB, there are no upper, upper left and upper right sub-blocks, and therefore, prediction modes 1 and 8 referring only to the left side, that is, the 0th sub-block SB, and an average value (0th Only the prediction mode 2 using the average value of the four pixels at the right end of the sub-block SB) can be used, and any prediction mode cannot be used unless the encoding of the 0th sub-block SB is finished. The same applies to the fifth sub-block SB.

  However, when the macro block MB constitutes the right adjacent portion at the top left of the screen, the 0th sub block SB is the fifth sub block SB that constitutes the top left macro block MB of the screen. You can refer to it.

  When the macro block MB constitutes the lower left upper part of the screen, the 0th sub-block SB corresponds to the 9th and 11th sub blocks constituting the upper left macro block MB of the screen. The block SB can be referred to, and the 1st sub-block SB can refer to the 9th, 11th and 14th sub-blocks SB constituting the upper left macroblock MB of the screen.

  Thus, in intra prediction, if encoding of the sub-block SB to be referred to is not completed, all nine prediction modes cannot be executed or all usable prediction modes are executed. Therefore, unless all the sub-blocks SB to be referenced are encoded, intra prediction cannot be started for the sub-block SB to be encoded, which is not efficient. .

  Therefore, the inventors consider that the efficiency can be improved if intra prediction can be performed at the stage where encoding for the immediately preceding sub-block SB is not finished, and the process until the optimum prediction mode is determined is divided into two stages. In the first stage, intra prediction using not only encoded pixel values but also pixel values before encoding (pixel values in an unencoded state) simply as prediction values (hereinafter referred to as simple intra prediction). In the second stage, the technical idea of performing the normal intra prediction for the narrowed prediction mode and determining the optimum prediction mode has been reached.

<Generation method of image data with simple intra prediction>
Hereinafter, a method for generating image data having simplified intra prediction will be described as an embodiment according to the present invention.

<Configuration and operation of encoder>
FIG. 12 shows an H.264 execution for generating image data having simple intra prediction. 2 is a block diagram illustrating a configuration of an H.264 encoder 100. FIG.

  As illustrated in FIG. 12, the encoder 100 includes a simple intra prediction unit 1, an intra prediction unit 2, a conversion unit 3, a quantization unit 4, an entropy encoding unit 5, an inverse quantization unit 6, an inverse conversion unit 7, and a loop filter 8. A frame memory 9, an inter-frame prediction unit 10, and a motion prediction unit 11.

  The input image signal input to the encoder 100 is given to the simple intra prediction unit 1 and the motion prediction unit 11 and is also used to take a difference from the predicted image output from the inter-frame prediction unit 10.

  The simplified intra prediction unit 1 uses not only the encoded pixel value of the reference target sub-block (first sub-block) but also the pixel value before encoding as the predicted value (first predicted value). A maximum of nine prediction modes are executed to obtain a prediction image (first prediction image) in each prediction mode, and the pixel value of the encoding target pixel of the encoding target sub-block (second sub-block) Get the absolute value sum of the differences. Then, by selecting some of the absolute value sums in descending order, the prediction mode is narrowed down to select the optimal prediction mode candidate.

  When there are few usable prediction modes, only one prediction mode may be selected as the optimum prediction mode candidate as a result.

  The result of narrowing down by the simple intra prediction unit 1 is given to the intra prediction unit 1 together with the input image signal, and the intra prediction unit 1 executes normal intra prediction for the optimum prediction mode candidate obtained by narrowing down. Determine the optimal prediction mode.

  Here, normal intra prediction is to perform prediction using an encoded pixel value as a prediction value (second prediction value). At this stage, encoding for the immediately preceding sub-block is finished. Therefore, there is no inconvenience in performing normal intra prediction. Note that simple intra prediction and intra prediction will be further described later.

  Difference data between each encoding target pixel and the predicted image (second predicted image) obtained by the optimal prediction mode determined in the intra prediction unit 2 is given to the conversion unit 3. Note that the conversion unit 3 is configured to be supplied with the difference data obtained in the above-described intraframe mode in which intraframe prediction is performed and the difference data obtained in the interframe mode in which interframe prediction is performed by switching between modes. ing.

  The transform unit 3 performs discrete cosine transform (DCT) on the given difference data and outputs transform coefficients. In H.264, integer transform (Integer Transform) in which floating point calculation is converted into an integer is used. The unit of conversion is a 4 × 4 pixel unit.

  The quantization unit 4 quantizes the DCT transform coefficient given from the transform unit 3 and outputs the quantized transform coefficient.

  The quantized transform coefficient is given to the entropy coding unit 5, and in the entropy coding unit 5, in the baseline profile and the X profile, an exponential Golomb code and a CAVLC (Context -Based Adaptive Variable Length Coding), and in the main profile and high profile, CABAC (Context-based Adaptive Binary Arithmetic Coding) is used to perform entropy coding on the quantized transform coefficients. The compressed image data is output as a bit stream.

  Note that the quantized transform coefficient is also supplied to the inverse quantization unit 6 and further returned to the difference data through the inverse transform unit 7.

  In the inter-frame mode, the switch SW1 is turned on and the difference data is added to the pixel value of the immediately preceding frame created by the inter-frame prediction unit 10 and then stored in the frame memory 9 via the loop filter 8. The The data of this frame is the same as the encoded image reproduced on the decoder side, and is used when the next frame is created as the immediately preceding frame.

  On the other hand, in the case of the intra-frame mode, the switch SW2 is turned on, and the difference data after inverse transformation is added to the pixel value of the predicted image obtained by the optimal prediction mode determined in the intra prediction unit 2, The encoded image data is given to the simple intra prediction unit 1 and the intra prediction unit 2 and is used for simple intra prediction and intra prediction. The encoded image data is the same as the image data reproduced on the decoder side.

  Here, the image data reproduced on the decoder side by the transformation process, the quantization process, the inverse transformation process and the inverse quantization process in the transformation unit 3, the quantization unit 4, the inverse quantization unit 6 and the inverse transformation unit 7 The process for obtaining the same image data is called local decoding.

  The loop filter 8 is a deblocking filter that smoothes only the block boundary of integer conversion and suppresses the generation of block noise.

  The inter-frame prediction unit 10 uses the image data of the immediately previous frame stored in the frame memory 9 to perform motion compensation that creates a predicted image of the next frame in consideration of screen motion. The predicted image created here is used to obtain a difference from the input image in the inter-frame mode.

  Further, the motion prediction unit 11 detects a motion vector from the input image signal, performs motion prediction of the screen, and gives the result to the inter-frame prediction unit 10.

<Pipeline processing>
Next, simple intra prediction and intra prediction will be described with reference to FIG. 12 and FIG. 13 schematically showing pipeline processing.

  In FIG. 13, the horizontal axis indicates the passage of time, and a series of processing cycles is shown as a block processing cycle. In FIG. 13, three block processing cycles (n−1), (n), and (n + 1) are shown. Show.

  Here, assuming that the n-th sub-block is a sub-block to be encoded, the simplified intra prediction unit 1 performs prediction modes 0 to 0 described with reference to FIGS. 3 to 11 for the n-th sub-block. Nine prediction modes of 8 are executed.

  In simple intra prediction, when the encoding of a sub-block to be referenced is not finished, the pixel value of the input image signal is used as a prediction value.

  For example, if the 13th subblock SB shown in FIG. 1 is the nth subblock, all the nine prediction modes are executed if the encoding of the 12th subblock SB is not completed in normal intra prediction. However, in the simple intra prediction, all the nine prediction modes can be executed by using the unencoded pixel of the 12th sub-block SB, that is, the pixel value of the encoding target pixel. .

  Further, for the first sub-block SB in FIG. 1, the prediction mode is used in a state in which the encoding of the 0-th sub-block SB is not finished by using the pixel value of the encoding target pixel of the 0-th sub-block SB. 1, 2 and 8 can be performed.

  As shown in FIG. 13, simple intra prediction for the nth sub-block is performed in the block processing cycle (n), but it takes time to execute all nine prediction modes, and the block processing cycle (n) The simple intra prediction is executed over almost the entire period.

  Then, by selecting some of the absolute value sums of the differences between the predicted image and the initial pixel value of the encoding target pixel UP obtained in each mode, the prediction mode is narrowed down by selecting some of them in ascending order. I do.

  As described above, the reason why the prediction mode is not determined to be one, but is narrowed down to several optimum prediction mode candidates is that the simple intra prediction is not as accurate as the normal intra prediction.

  Further, in the block processing cycle (n), in parallel with the simple intra prediction, the intra prediction unit 2 executes normal intra prediction on the (n−1) th subblock.

  For the (n−1) th sub-block, since the prediction mode is narrowed down by executing the simple intra prediction in the block processing cycle (n−1), the result of the narrowing down is reflected and the block processing cycle (n ), The optimum prediction mode can be determined within the processing cycle period.

  That is, in the intra prediction performed here, only the optimal prediction mode candidates narrowed down by the simple intra prediction need be executed again, and the optimal prediction mode can be set in a shorter time than the simple intra prediction. Can be determined. Moreover, in the intra prediction performed here, since all reference object pixels are encoded, prediction with high accuracy is possible.

  In the block processing cycle (n−1), since local decoding for the n−2th subblock has been completed, it is difficult to perform normal intra prediction for the n−1th subblock. Does not occur.

  After performing the normal intra prediction and determining the optimal prediction mode for the n-1st sub-block, in the block processing cycle (n-1), the optimal prediction mode is used, and n-1 Perform local decoding of sub-block number.

  The pixel data of the (n−1) th subblock encoded by local decoding is stored in the memory in the intra prediction unit 2 and used in normal intra prediction for the nth subblock.

  Similarly, in the block processing cycle (n + 1), the simple intra prediction for the n + 1-th subblock is executed, and the normal prediction is executed for the n-th subblock to determine the optimum prediction mode. Thereafter, the local decoding of the nth sub-block is executed using the difference data obtained in the optimum prediction mode.

  FIG. 14 is a flowchart showing the flow of simple intra prediction, normal intra prediction, and local decoding to be performed on the encoding target sub-block.

  In FIG. 14, when data of an encoding target image is input to the encoder 100, simple intra prediction is performed in the simple intra prediction unit 1 (step S1), and then the predicted image obtained in each mode, and the encoding are performed. The prediction mode is narrowed down by selecting several from the sum of absolute values of differences from the original pixel value of the target pixel in ascending order of values (step S2). The intra prediction unit 2 is also provided with the data of the encoding target pixel.

  Then, the intra prediction unit 2 performs intra prediction on the optimal prediction mode candidates narrowed down in step S2 using the encoded pixel values (step S3), and then obtained in each mode. The optimum prediction mode is determined by selecting one of the absolute value sums of differences between the prediction image (first prediction image) and the original pixel value of the encoding target pixel to select one having the smallest value. (Step S4).

  Next, orthogonal transform (for example, DCT) is performed on the difference data between the predicted image (second predicted image) obtained by the determined optimal prediction mode and the original pixel value of each encoding target pixel. (Step S5), followed by quantization in the quantization unit 4 (step S6), inverse quantization in the inverse quantization unit 6 (step S7), and inverse transformation in the inverse transformation unit 7 (step S8).

  Then, encoded image data is generated by adding data (difference data) after inverse transformation to the predicted image obtained in the optimal prediction mode, and fed back to the intra prediction unit 2.

<Effect>
As described above, in the encoder 100, the simplified intra prediction unit 1 that performs intra prediction using the pixel value before encoding as a prediction value is provided, and the image is processed even when the encoding for the immediately preceding sub-block is not finished. Prediction can be started, the time spent for intra prediction can be shortened, and the efficiency of intra prediction can be improved.

  In addition, simple intra prediction and normal intra prediction for the same encoding target sub-block are processed in different processing cycles by pipeline processing, and local decoding is performed after normal intra prediction. Data encoding processing can be performed efficiently.

It is a figure which represents typically the macroblock comprised by 16x16 pixel. It is a figure which represents typically the 16 encoding object pixel which comprises a subblock, and the reference pixel used in intra prediction. It is a figure which represents prediction mode 0 typically. It is a figure which represents prediction mode 1 typically. It is a figure which represents the prediction mode 2 typically. It is a figure which represents the prediction mode 3 typically. It is a figure which represents the prediction mode 4 typically. It is a figure which represents the prediction mode 5 typically. It is a figure which represents the prediction mode 6 typically. It is a figure which represents the prediction mode 7 typically. It is a figure which represents the prediction mode 8 typically. A method for generating image data having simplified intra prediction according to the present invention is described. 2 is a block diagram illustrating a configuration of an H.264 encoder. FIG. It is a figure which shows a pipeline process typically. It is a flowchart explaining the production | generation method of the image data which has the simple intra prediction which concerns on this invention.

Explanation of symbols

MB Macro block SB Sub block EP Encoded pixel UP Encoding target pixel BL block

Claims (4)

A method of generating image data having intra prediction that performs image prediction of a second sub-block to be encoded using pixel values of a plurality of encoded first sub-blocks,
Prior to the intra prediction,
When the encoding of the first sub-block is not completed, the pixel value before the encoding of the first sub-block is set as a first predicted value, and the first predicted value is used to Perform interpolation for each of a plurality of prediction modes for two sub-blocks,
Based on the result of interpolation for each of the plurality of prediction modes, having at least one of the plurality of prediction modes to be a simple intra prediction as an optimal prediction mode candidate,
The intra prediction performs the image prediction using the at least one optimal prediction mode candidate,
The intra prediction is performed after encoding is completed for all of the first sub-blocks;
The pixel value after encoding of the first sub-block that has been encoded is set as a second predicted value, and the at least one optimal value is used for the second sub-block using the second predicted value. Perform interpolation for each prediction mode candidate,
On the basis of the results of at least one optimum prediction mode candidate for each of the interpolation, it has rows the picture prediction to determine the optimum prediction mode from the at least one optimum prediction mode candidates,
In the simple intra prediction ,
For the first sub-block that has been encoded , image data that uses the pixel value after encoding of the first sub-block or the pixel value before encoding as the first predicted value Generation method. In the simple intra prediction ,
The absolute value sum of the differences between the first predicted image of the second sub-block obtained as a result of the interpolation for each of the plurality of prediction modes and the pixel value before the interpolation of the second sub-block is executed The image data generation method according to claim 1, wherein the at least one optimum prediction mode candidate is selected by acquiring for each of the plurality of prediction modes and selecting in order from the smallest absolute value sum . The intra prediction is
The difference between the second predicted image of the second sub-block obtained as a result of interpolation for each of the at least one optimum prediction mode candidate and the pixel value before the interpolation of the second sub-block is executed The method of generating image data according to claim 1 , wherein an absolute value sum is obtained for each of the at least one optimum prediction mode candidate, and the optimum prediction mode is determined by selecting the smallest absolute value sum. . The simple intra prediction and the intra prediction are executed by pipeline processing,
The pipeline processing is
In the nth processing cycle in which the simple intra prediction is executed for the nth subblock, for the n−1th subblock in which the simple intra prediction is executed in the n−1th processing cycle. 2. The image data according to claim 1 , wherein the intra prediction is performed and local decoding is performed on the n−1th sub-block using the optimal prediction mode determined by the intra prediction. Generation method.

Images