JP6324590B2 - Image encoding device, image encoding method and program, image decoding device, image decoding method and program - Google Patents

Description

  The present invention relates to an image encoding device, an image encoding method and program, an image decoding device, an image decoding method and a program.

  As an encoding method for compression recording of moving images, H.264 is used. H.264 / MPEG-4 AVC (hereinafter abbreviated as H.264) is known. (Non-Patent Document 1) H.264 defines a plurality of profiles that define encoding technology limitations. For example, the High 10 profile corresponds to an image having a bit depth of 8 bits to 10 bits.

  In recent years, H.C. As a successor to H.264, activities to start international standardization of a more efficient coding method have started. JCT-VC (Joint Collaborative Team on Video Coding) was established between ISO / IEC and ITU-T. In this JCT-VC, standardization is underway as a HEVC (High Efficiency Video Coding) coding method (hereinafter abbreviated as HEVC).

  Also in HEVC, a Main 10 profile corresponding to an image having a bit depth of 8 bits to 10 bits is defined. (Non-Patent Document 2)

ITU-TH. H.264 (06/2011) Advanced video coding for generic audiovisual services JCT-VC contribution JCTVC-K1003_v10. doc Internet <http: // phenix. int-evry. fr / jct / doc_end_user / documents / 11_Shanghai / wg11 />

  HEVC has a configuration that emphasizes ease of mounting by reducing calculation accuracy in accordance with the bit depth of an image in processing such as orthogonal transformation and motion compensation. For example, the following formula (1) is one of calculation formulas used for motion compensation processing of sub-pixels in motion compensation of color difference signals.

_{ab 0,0 = (- 2 × B} -1,0 + 58 × B 0,0 + 10 × B 1,0 -2 × B 2,0) >> shift1 ... (1)
(However, shift1 = chrominance bit depth−8, and “>>” represents a bit shift to the right.)
In the above equation (1), B _{i, j} represents the color difference pixel at the integer pixel position, and ab _0,0 represents the intermediate value for calculating the color difference pixel at the decimal pixel position. Since the expression (1) always includes bit shift processing to the right by shift1 depending on the bit depth, the range of values that the intermediate values ab _0,0 can take is constant regardless of the bit depth of the image. Since such arithmetic processing is introduced, it is considered that the implementation cost of hardware does not increase so much in HEVC even when an image with a higher bit depth is supported. On the other hand, the calculation represented by the above-described bit shift processing causes a problem that the calculation accuracy is lowered for an image with a high bit depth and the image quality is not improved.

  Therefore, the present invention has been made to solve the above-described problems. In other words, both encoding processing that maintains constant accuracy even at high bit depths regardless of the bit depth of the image, and encoding processing that emphasizes ease of implementation by reducing the calculation accuracy at high bit depths depending on the bit depth. The purpose is to realize supported encoding / decoding.

In order to solve the above-described problems, the image coding apparatus according to the present invention has the following configuration, that is, an image coding apparatus that codes an image, and performs conversion processing on a difference between a target pixel and a prediction pixel A code that encodes the derivation means for deriving the coefficient by performing data, data corresponding to the coefficient, and a flag indicating whether the minimum and maximum values related to the coefficient are determined by the bit depth or fixed The image decoding apparatus of the present invention has the following configuration. That is, an image decoding apparatus for decoding an image from a bitstream, wherein data corresponding to a coefficient related to a pixel of the image and a minimum value and a maximum value related to the coefficient are determined by a bit depth or fixed. And a deriving unit for deriving a difference between the target pixel and the predicted pixel based on the coefficient by performing a conversion process.

  According to the present invention, an encoding process that maintains a constant accuracy even at a high bit depth regardless of the bit depth of the image, and an encoding process that emphasizes ease of implementation by reducing the calculation accuracy at a high bit depth depending on the bit depth. Encoding / decoding that supports both of the above can be realized. As a result, these encoding processes can be switched according to the required specifications for each application.

1 is a block diagram illustrating a configuration of an image encoding device according to a first embodiment. FIG. 3 is a block diagram illustrating a configuration of an image decoding device according to a second embodiment. FIG. 9 is a block diagram illustrating a configuration of an image encoding device according to a third embodiment. FIG. 9 is a block diagram illustrating a configuration of an image decoding device according to a fourth embodiment. 7 is a flowchart showing image encoding processing in the image encoding apparatus according to the first embodiment. 7 is a flowchart showing image decoding processing in the image decoding apparatus according to the second embodiment. 10 is a flowchart showing image encoding processing in the image encoding apparatus according to the third embodiment. 10 is a flowchart showing image decoding processing in the image decoding apparatus according to the fourth embodiment. FIG. 7 is a block diagram illustrating a configuration of an image encoding device according to a fifth embodiment. FIG. 9 is a block diagram illustrating a configuration of an image decoding device according to a sixth embodiment. FIG. 10 is a block diagram showing another configuration of the image encoding device according to the fifth embodiment. FIG. 10 is a block diagram showing another configuration of the image decoding apparatus according to the sixth embodiment. 10 is a flowchart showing image encoding processing in the image encoding device according to the fifth embodiment. 10 is a flowchart showing image decoding processing in the image decoding apparatus according to the sixth embodiment. Another flowchart showing an image encoding process in the image encoding device according to the fifth embodiment. Another flowchart showing an image decoding process in the image decoding apparatus according to the sixth embodiment. The figure which shows an example of the bit stream structure produced | generated by Embodiment 1 and decoded by Embodiment 2. The figure which shows an example of the bit stream structure produced | generated by Embodiment 3 and decoded by Embodiment 4. The figure which shows an example of the bit stream structure produced | generated by Embodiment 5 and decoded by Embodiment 6. 1 is a block diagram showing an example of a hardware configuration of a computer applicable to an image encoding device and a decoding device of the present invention. The figure which shows the relationship between the range information in Embodiment 1 and Embodiment 2, the bit depth of an image, and the range which a quantization coefficient can take

  Hereinafter, the present invention will be described in detail based on the preferred embodiments with reference to the accompanying drawings. The configurations shown in the following embodiments are merely examples, and the present invention is not limited to the illustrated configurations.

<Embodiment 1>
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an image encoding apparatus according to this embodiment. In FIG. 1, reference numeral 101 denotes a terminal for inputting image data.

  An input unit 102 analyzes the bit depth of the input image data and divides it into square blocks. Reference numeral 103 denotes a transform quantization operation accuracy information generation unit that generates transform quantization operation accuracy selection information described later. At the same time, transform quantization calculation accuracy information indicating the calculation accuracy of the transform quantization process used in the transform quantization unit 106 and the inverse quantization inverse transform process used in the inverse quantization inverse transform unit 107 is generated. Reference numeral 104 denotes a header encoding unit that encodes information necessary for decoding a bit stream including bit depth information of an image to generate header code data.

  A prediction unit 105 refers to the frame memory 109 in units of blocks divided into squares, performs intra prediction that is intra-frame prediction, inter prediction that is inter-frame prediction, and the like. Generate an error. Reference numeral 106 denotes a transform quantization unit, which orthogonally transforms the prediction error generated by the prediction unit 105 in units of blocks to calculate transform coefficients, and further quantizes the transform coefficients to calculate quantization coefficients. Reference numeral 107 denotes an inverse quantization inverse transform unit that inversely quantizes the quantization coefficient generated by the transform quantization unit 106 to reproduce the transform coefficient, and further performs inverse orthogonal transform to reproduce a prediction error.

  An image reproduction unit 108 refers to the frame memory 109 based on the prediction information generated by the prediction unit 105 and performs intra prediction, inter prediction, and the like. From the prediction error generated by the inverse quantization and inverse conversion unit 107 Generate a playback image. Reference numeral 109 denotes a frame memory that holds an image reproduced by the image reproducing unit 108. A block encoding unit 110 encodes the prediction information generated by the prediction unit 105 and the quantization coefficient generated by the transform quantization unit 106 to generate block code data. Reference numeral 111 denotes an integrated encoding unit that forms a bit stream from the header code data and block code data generated in the previous stage and outputs the bit stream. Reference numeral 112 denotes a terminal that outputs the bitstream generated by the integrated encoding unit 111 to the outside.

  An image encoding operation in the image encoding apparatus will be described below. In the present embodiment, moving image data is input in units of frames, but still image data for one frame may be input.

  The image data for one frame input from the terminal 101 is input to the input unit 102. In this embodiment, it is assumed that 10-bit depth image data is input, but the bit depth of the input image data is not limited to this. The input unit 102 analyzes the bit depth of the input image data, and outputs the bit depth information to the subsequent range information generation unit 103 and the header encoding unit 104. However, it is also possible to adopt a configuration in which the bit depth information is separately given from the outside and input to the transform quantization calculation accuracy information generation unit 103 and the header encoding unit 104, respectively. The input image data is divided into square blocks and output to the prediction unit 105.

  The transform quantization calculation accuracy information generation unit 103 uses a transform quantization process that adjusts the calculation precision according to the bit depth and prioritizes mounting ease, or uses a transform quantization process that fixes the calculation precision regardless of the bit depth. And the information is used as transform quantization calculation accuracy selection information. Hereinafter, the former transform quantization processing in which the calculation accuracy is adjusted according to the bit depth is referred to as implementation-oriented transform quantization processing, and the latter transform quantization processing in which the calculation accuracy is fixed is referred to as accuracy-oriented transform quantization processing. In the present embodiment, when the former implementation-oriented transform quantization process is selected, the transform quantization operation accuracy selection information is 0, and when the latter accuracy-oriented transform quantization process is selected, the transform quantization operation is performed. It is assumed that the accuracy selection information is 1. However, the combination of the selected transform quantization process and transform quantization calculation accuracy selection information is not limited to these. Also, the method for determining the transform quantization calculation accuracy selection information is not particularly limited, and may be determined prior to the encoding process assuming an application in which the present encoding device and the corresponding decoding device are used. It may be selected by a user (not shown). For example, when it is assumed that the encoding apparatus according to the present embodiment is used in an application in which calculation accuracy is important, the transform quantization calculation accuracy selection information is set to 1, otherwise it is set to 0. .

  Next, the transform quantization operation accuracy information generation unit 103 generates transform quantization operation accuracy information based on the above-described transform quantization operation accuracy selection information and the bit depth information input from the input unit 102. When the transform quantization operation accuracy selection information is 1, the difference value between the bit depth of the image and the reference bit depth of 8 bits is used as transform quantization operation accuracy information. In this embodiment, since the bit depth of the image is 10 bits, the transform quantization calculation accuracy information is 2. When the transform quantization operation accuracy selection information is 0, 0 is used as transform quantization operation accuracy information. However, the combination of value and meaning of transform quantization calculation accuracy information is not limited to the above, and indicates that the calculation accuracy of transform quantization processing is improved when the bit depth of the image is larger than the reference bit depth. Any information can be used.

  The generated transform quantization operation accuracy selection information is output to the header encoding unit 104, and the transform quantization operation accuracy information is output to the transform quantization unit 106 and the inverse quantization inverse transform unit 107.

The header encoding unit 104 encodes information necessary for decoding including the bit depth information input from the input unit 102 and the transform quantization operation accuracy selection information input from the transform quantization operation accuracy information generation unit 103. The header code data is generated. This header code data corresponds to the header portion of the bit stream. The generated header code data is output to the integrated encoding unit 111, while the prediction unit 105 receives the block-unit image data divided by the input unit 102 and stores the encoded image data stored in the frame memory 109. Prediction in units of blocks is performed with reference to the pixels as appropriate, and a predicted image is generated. A prediction error is generated as a difference between the input image of the block unit and the predicted image, and is input to the transform quantization unit 106. Further, the prediction unit 105 generates information necessary for prediction such as a motion vector and a prediction mode as prediction information, and outputs the prediction information to the image reproduction unit 108 and the block encoding unit 110.

  The transform quantization unit 106 first inputs transform quantization operation accuracy information from the transform quantization operation accuracy information generation unit 103, and determines the operation accuracy in the transform quantization process. In the present embodiment, based on the table shown in FIG. 21, the range that can be taken by each calculation result such as one-dimensional orthogonal transformation or quantization processing in the horizontal and vertical directions is determined as the calculation accuracy. However, the combination of the transform quantization calculation accuracy information and the range that each calculation result can take is not limited to these. In this embodiment, since transform quantization calculation accuracy information having a value of 0 or 2 is input, each calculation result has a range of −32768 to 32767 or −131072 to 131071, respectively. There is no particular limitation on the processing when each calculation result falls outside the above range, but the result can be kept within the above range by clip processing or bit shift processing.

  Next, based on the above-described determined calculation accuracy, orthogonal transform is performed on the prediction error input from the prediction unit 105, a transform coefficient is generated, the transform coefficient is further quantized, and a quantized coefficient is generated. . The generated quantization coefficient is output to the inverse quantization inverse transform unit 107 and the block coding unit 110.

  Similarly to the transform quantization unit 106, the inverse quantization inverse transform unit 107 first inputs the transform quantization computation accuracy information from the transform quantization computation accuracy information generation unit 103, and increases the computation accuracy in the inverse quantization inverse transform process. decide. In the present embodiment, similar to the transform quantization unit 106, based on the table shown in FIG. 21, each computation result such as inverse quantization processing and one-dimensional orthogonal transform processing in each of the vertical and horizontal directions is calculated as computation accuracy. The possible range shall be determined.

  Next, based on the determined calculation accuracy, the quantization coefficient input from the transform quantization unit 106 is inversely quantized to reproduce the transform coefficient, and further, inverse orthogonal transform to the reconstructed transform coefficient is performed for prediction. Play back the error. The reproduced prediction error is output to the image reproduction unit 108.

  In the image reproduction unit 108, based on the prediction information input from the prediction unit 105, a predicted image is generated by referring to the frame memory 109 as appropriate, and the generated prediction image and the inverse quantization inverse transform unit 107 are input. A reproduced image is generated from the prediction error. The generated reproduced image is output to the frame memory 109 and stored.

  The block encoding unit 110 entropy-encodes the quantization coefficient input from the transform quantization unit 106 and the prediction information input from the prediction unit 105 in units of blocks to generate block code data. The entropy encoding method is not particularly specified, but Golomb encoding, arithmetic encoding, Huffman encoding, or the like can be used. The generated block code data is output to the integrated encoding unit 111.

  The integrated encoding unit 111 multiplexes the header code data generated and input by the header encoding unit 104 and the block code data input from the block encoding unit 110 prior to the block-by-block encoding process to generate a bitstream. Is formed. Finally, the bit stream formed by the integrated encoding unit 111 is output from the terminal 112 to the outside.

  FIG. 17A shows an example of the bit stream generated in this embodiment. The transform quantization operation accuracy selection information is included as a transform quantization operation accuracy selection information code in one of headers such as a sequence and a picture. Similarly, bit depth information is also included in one of the headers as a bit depth information code.

  However, the configuration of the bit stream is not limited to this, and as shown in FIG. 17B, instead of encoding the transform quantization calculation accuracy selection information code, the corresponding profile is determined, and the determined profile is It may be encoded as a profile information code. For example, it is assumed that there are a main 10-bit profile and a main 10-bit high-accuracy profile, each corresponding to transform quantization calculation accuracy selection information = 0 and transform quantization calculation accuracy selection information = 1. That is, in the main 10-bit profile, the range that can be taken by each operation result of the transform quantization processing is constant regardless of the bit depth of the image, and in the main 10-bit high-accuracy profile, the above range varies depending on the bit depth of the image. It will be. In this case, when the transform quantization calculation accuracy selection information is 0, a code indicating the main 10-bit profile is encoded as a profile information code. I do not care.

  FIG. 5 is a flowchart illustrating image encoding processing in the image encoding device according to the first embodiment.

  First, in step S501, the input unit 102 analyzes the bit depth of input image data and generates bit depth information. In step S502, the transform quantization operation accuracy information generation unit 103 generates transform quantization operation accuracy selection information for selecting transform quantization operation accuracy information indicating the operation accuracy in the transform quantization process. In step S503, the transform quantization operation accuracy information generation unit 103 generates transform quantization operation accuracy information based on the transform quantization operation accuracy selection information generated in step S502 and the bit depth information generated in step S501. To do. In step S504, the header encoding unit 104 encodes information necessary for decoding such as the bit depth information generated in step S501 and the transform quantization calculation accuracy selection information generated in step S502 to generate a header code. Generate data.

  In step S505, the integrated encoding unit 111 forms the header part of the bit stream from the header code data generated in step S504, and outputs it. In step S506, the input unit 102 cuts out a square block from the input image data, and the prediction unit 105 performs block unit prediction on the cut out block unit image data to generate a predicted image. A prediction error is generated as the difference between the input image data in block units and the predicted image. Furthermore, information necessary for prediction such as a motion vector and a prediction mode is generated as prediction information.

  In step S507, the transform quantization unit 106 first determines the computation accuracy in the transform quantization process based on the transform quantization computation accuracy information generated in step S503. Then, based on the determined calculation accuracy, orthogonal transform is performed on the prediction error generated in step S506 to generate a transform coefficient, and the generated transform coefficient is quantized to generate a quantized coefficient. In step S508, the inverse quantization inverse transform unit 107 first determines the computation accuracy in the inverse quantization / inverse transformation process based on the transform quantization computation accuracy information generated in step S503, as in step S507. Then, based on the determined calculation accuracy, the quantization coefficient generated in step S507 is inversely quantized to reproduce the transform coefficient, and further the inverse orthogonal transform to the reproduced transform coefficient is performed to reproduce the prediction error.

  In step S509, the image reproduction unit 108 generates a predicted image by appropriately referring to the frame memory 109 based on the prediction information generated in step S506. Then, a reproduced image is generated from the generated predicted image and the prediction error reproduced in step S508 and stored in the frame memory 109. In step S510, the block encoding unit 110 encodes the prediction information generated in step S506 and the quantization coefficient generated in step S507 to generate block code data. In addition, the integrated encoding unit 111 generates a bit stream including other code data. In step S511, the image encoding apparatus determines whether or not encoding of all the blocks in the frame has been completed. If the encoding has ended, the encoding process ends. If not, the next block is determined. The process returns to step S506 as a target.

  With the above configuration and operation, particularly by encoding the transform quantization calculation accuracy selection information in step S504, a bitstream that can be switched between encoding processes with different calculation accuracy and mounting cost according to the required specifications of the application is generated. can do.

  In the present embodiment, the flow of the encoding process has been described in the order of steps S508, S509, and S510. However, the order is not limited to this, and step S510 may be executed after step S507.

  In the present embodiment, only the transform quantization processing in steps S507 and S508 is changed based on the transform quantization calculation accuracy selection information. However, as the quantization coefficient range is changed, the code in step S510 is changed. The configuration may also be such that the digitization process is also changed. In this case, the transform quantization operation accuracy selection information or the transform quantization operation accuracy information is also input to the block coding unit 110. In this case, since an optimal entropy encoding method can be selected according to the range of quantization coefficients, more efficient encoding can be realized.

  Further, when the bit depth of the image data to be encoded is 8 bits, the transform quantization calculation accuracy selection information code can be omitted. That is, when the bit depth is 8 bits, the transform quantization calculation accuracy information is uniquely determined to be 0, and redundant codes can be reduced.

Further, in the present embodiment, based on the table shown in FIG. 21, the range that can be taken by each calculation result such as one-dimensional orthogonal transformation and quantization processing in the horizontal and vertical directions is determined as the calculation accuracy. It is not limited to. For example, the conversion quantization unit 106 when the transform quantization operation precision information was aq, may be calculated an arithmetic accuracy as ^{-2 (15 + aq) ~2 (} 15 + aq) -1.

  In the main 10-bit profile and the main 10-bit high-precision profile, the transform quantization calculation accuracy selection information is always set to 0 in the former, and the transform quantization calculation accuracy selection information code is set to 0/1 in the latter. You may use the structure to select. As a result, the calculation accuracy can be selected even with a high-precision profile.

  In this embodiment, in step S508 of FIG. 5, the calculation accuracy in the inverse quantization / inverse transform process is determined based on the transform quantization calculation accuracy information generated in step S503. However, the calculation accuracy obtained in step S507 is determined. Of course you can use.

  In addition, as shown in FIG. 17A, the bit stream generated in the present embodiment is encoded in the order of the transform quantization calculation accuracy selection information code and the bit depth information code. It is not limited.

<Embodiment 2>
FIG. 2 is a block diagram showing the configuration of the image decoding apparatus according to Embodiment 2 of the present invention. In the present embodiment, description will be made by taking the decoding of the bitstream generated in the first embodiment as an example.

  Reference numeral 201 denotes a terminal for inputting a bit stream. A separation / decoding unit 202 separates the bitstream into header code data, which is information related to decoding processing, and block code data, which is block unit information such as quantization coefficients and prediction information, and outputs it to the subsequent stage. A header decoding unit 203 decodes the above-described header code data and reproduces information related to the decoding process. Reference numeral 204 denotes a transform quantization calculation accuracy information setting unit that generates transform quantization calculation accuracy information indicating the calculation accuracy of the inverse quantization inverse transform process used by the inverse quantization inverse transform unit 206. Reference numeral 205 denotes a block decoding unit that decodes block code data and reproduces quantization coefficients and prediction information.

  Reference numeral 206 denotes an inverse quantization inverse transform unit that inversely quantizes the quantization coefficient reproduced by the block decoding unit 205 to reproduce a transform coefficient, and further performs inverse orthogonal transform to reproduce a prediction error. Reference numeral 207 denotes an image reproduction unit, which performs intra prediction, inter prediction, and the like with reference to the frame memory 208 based on the prediction information reproduced by the block decoding unit 205, and generates a prediction error generated by the inverse quantization inverse transformation unit 206 Reproduced image data is generated from A frame memory 208 holds image data reproduced by the image reproduction unit 207. A terminal 209 outputs the reproduced image data to the outside.

  An image decoding operation in the image decoding apparatus will be described below. In the present embodiment, the bit stream generated in the first embodiment is decoded.

  In FIG. 2, the bit stream input from the terminal 201 is input to the separation / decoding unit 202. In the present embodiment, it is assumed that the bit stream shown in FIG. The separation / decoding unit 202 separates the input bitstream into header code data that is information related to decoding processing and block code data that is information in block units, and outputs the result to the subsequent stage. The header code data is output to the header decoding unit 203, and the block code data is output to the block decoding unit 205. The header decoding unit 203 decodes information necessary for decoding from the header code data input from the separation decoding unit 202, and reproduces transform quantization calculation accuracy selection information and bit depth information. The reproduced transform quantization operation accuracy selection information and bit depth information are output to the transform quantization operation accuracy information setting unit 204.

  The transform quantization operation accuracy information setting unit 204 generates transform quantization operation accuracy information based on the transform quantization operation accuracy selection information and the bit depth information input from the header decoding unit 203. Similar to the transform quantization operation accuracy information generation unit 103 of the first embodiment, in this embodiment, when the transform quantization operation accuracy selection information is 1, the bit depth information and the reference bit depth of 8 bits are Is used as transform quantization calculation accuracy information. Since the bit stream generated in the first embodiment is obtained by encoding a 10-bit image, the bit depth information in this embodiment is also 10 bits, and thus the transform quantization calculation accuracy information is 2. On the other hand, when the transform quantization operation accuracy selection information is 0, 0 is used as transform quantization operation accuracy information. However, as in the first embodiment, the combination of transform quantization computation accuracy selection information and transform quantization computation accuracy information is not limited to these. The generated transform quantization calculation accuracy information is output to the inverse quantization inverse transform unit 206.

  On the other hand, the block decoding unit 205 decodes the block code data input from the separation decoding unit 202 and reproduces the quantization coefficient and the prediction information. The reproduced quantization coefficient is output to the inverse quantization and inverse transform unit 206, and the prediction information is output to the image reproduction unit 207. In the inverse quantization inverse transform unit 206, first, similarly to the inverse quantization inverse transform unit 107 of the first embodiment, based on the transform quantization operation accuracy information input from the transform quantization operation accuracy information setting unit 204, the inverse quantization is performed. The calculation accuracy in the inverse conversion process is determined. Also in this embodiment, similarly to the inverse quantization inverse transform unit 107 of the first embodiment, based on the table shown in FIG. 21, the inverse quantization processing and the one-dimensional orthogonal transform in each of the vertical and horizontal directions are performed as the calculation accuracy. It is assumed that the range that each arithmetic processing such as processing can take is determined.

  Further, the inverse quantization inverse transform unit 206 reproduces the transform coefficient by inversely quantizing the quantized coefficient input from the block decoding unit 205 based on the above-described determined calculation accuracy, and further reproduces the transformed transform coefficient. The prediction error is reproduced by inverse orthogonal transformation to. The reproduced prediction error is output to the image reproduction unit 207.

  In the image reproduction unit 207, based on the prediction information input from the block decoding unit 205, the frame memory 208 is appropriately referred to generate a prediction image, and the generated prediction image and the inverse quantization inverse transformation unit 206 are input. A reproduced image is generated from the predicted error. The reproduced image data is output to and stored in the frame memory 208. At the same time, the reproduced image data is output from the terminal 209 to the outside.

  FIG. 6 is a flowchart illustrating an image decoding process in the image decoding apparatus according to the second embodiment.

  First, in step S601, the separation / decoding unit 202 separates header code data, which is information related to decoding processing, from the input bitstream. In step S602, the header decoding unit 203 decodes information necessary for decoding from the header code data separated in step S601, and reproduces transform quantization calculation accuracy selection information and bit depth information. In step S603, the transform quantization operation accuracy information setting unit 204 generates transform quantization operation accuracy information based on the transform quantization operation accuracy selection information and the bit depth information reproduced in step S602. In step S604, the inverse quantization inverse transform unit 206 determines the computation accuracy in the inverse quantization / inverse transformation process based on the transform quantization computation accuracy information generated in step S603. In step S605, the block decoding unit 205 decodes the block code data separated as encoded data in units of blocks from the bit stream by the separation decoding unit 202, and reproduces the quantization coefficient and the prediction information.

  In step S606, the inverse quantization inverse transform unit 206 reproduces the transform coefficient by dequantizing the quantization coefficient generated in step S605 based on the calculation accuracy determined in step S604, and further reproduces the transformed transform. The prediction error is reproduced by inverse orthogonal transform to coefficients. In step S607, the image reproduction unit 207 generates a predicted image by appropriately referring to the frame memory 208 based on the prediction information reproduced in step S605. Then, a reproduced image is generated from the generated predicted image and the prediction error reproduced in step S606 and stored in the frame memory 208. At the same time, the reproduced image data is output from the terminal 209 to the outside. In step S608, the image decoding apparatus determines whether or not the decoding of all the blocks in the frame has been completed. If the decoding has been completed, the decoding process is terminated. Otherwise, the step is performed on the next block. The process returns to S605.

  With the above configuration and operation, decoding can be performed with different calculation accuracy and mounting cost according to the required specifications of the application generated in the first embodiment, particularly by decoding the transform quantization calculation accuracy selection information in step S602. The bitstream can be decoded.

  In the present embodiment, the input bit stream is obtained by independently encoding the transform quantization calculation accuracy selection information shown in FIG. 17A, but the present invention is not limited to this. For example, as shown in FIG. 17B, a profile information code indicating a corresponding profile may be encoded instead of encoding the transform quantization calculation accuracy selection information code. In this case, the transform quantization operation accuracy information setting unit 204 generates transform quantization operation accuracy information from the profile information code and the bit depth information.

  In the present embodiment, only the inverse quantization and inverse transform process in step S606 is changed based on the transform quantization calculation accuracy selection information. However, in accordance with the change of the quantization coefficient range, the decoding in step S605 is performed. It is good also as a structure which changes a process. In that case, the transform quantization operation accuracy selection information or the transform quantization operation accuracy information is also input to the block decoding unit 205, and the decoding process of the block decoding unit 205 is performed by the block encoding unit 110 of the first embodiment. It is necessary to support encoding processing. In this case, since an optimal entropy decoding method can be selected according to the range of quantization coefficients, it is possible to realize decoding of a bit stream encoded with higher efficiency.

<Embodiment 3>
FIG. 3 is a block diagram showing an image encoding apparatus according to this embodiment. In FIG. 3, the same numbers are assigned to portions that perform the same functions as those in FIG. 1 of the first embodiment, and description thereof is omitted.

  A motion compensation calculation accuracy information generation unit 323 generates motion compensation calculation accuracy selection information described later. At the same time, motion compensation calculation accuracy information indicating the calculation accuracy of the motion compensation processing used in the prediction unit 305 is generated. Reference numeral 304 denotes a header encoding unit that encodes information necessary for decoding a bitstream such as image bit depth information to generate header code data. It differs from the header encoding unit 104 of the first embodiment in that it encodes motion compensation calculation accuracy selection information, which will be described later, instead of transform quantization calculation accuracy selection information.

  A prediction unit 305 refers to the frame memory 109 in units of blocks divided into squares, performs intra prediction that is intra-frame prediction, inter prediction that is inter-frame prediction, and the like. Generate an error. The prediction unit 105 of the first embodiment is different in that motion compensation calculation accuracy information is input and inter prediction is performed based on the input motion compensation calculation accuracy information.

  Reference numeral 306 denotes a transform quantization unit that computes transform coefficients by orthogonally transforming the prediction error generated by the prediction unit 305 in units of blocks, and further quantizes the transform coefficients to calculate quantization coefficients. It differs from the transform quantization unit 106 of the first embodiment in that transform quantization processing accuracy information is not input and transform quantization processing is always performed with the same computation accuracy.

  Reference numeral 307 denotes an inverse quantization inverse transform unit that inversely quantizes the quantization coefficient generated by the transform quantization unit 306 to reproduce the transform coefficient, and further performs inverse orthogonal transform to reproduce a prediction error. It differs from the inverse quantization inverse transform unit 107 of the first embodiment in that the transform quantization computation accuracy information is not input and the inverse quantization inverse transform processing is always performed with the same computation accuracy.

  Reference numeral 308 denotes an image reproduction unit that performs intra prediction, inter prediction, and the like with reference to the frame memory 109 based on the prediction information generated by the prediction unit 306, and uses the prediction error generated by the inverse quantization inverse conversion unit 307. Generate a playback image. The difference from the image reproduction unit 108 of Embodiment 1 is that motion compensation calculation accuracy information is input and inter prediction is performed based on the input motion compensation calculation accuracy information.

  An image encoding operation in the image encoding apparatus will be described below.

  The input unit 102 analyzes the bit depth of the input image data and outputs the bit depth information to the subsequent motion compensation calculation accuracy information generation unit 323 and the header encoding unit 304. However, it is also possible to adopt a configuration in which the bit depth information is separately provided from the outside and input to the motion compensation calculation accuracy information generation unit 323 and the header encoding unit 304, respectively. Further, the input image data is divided into square blocks and output to the prediction unit 305.

  The motion compensation calculation accuracy information generation unit 323 determines whether to use a motion compensation process that adjusts the calculation accuracy according to the bit depth and prioritizes mounting ease, or a motion compensation process that fixes the calculation accuracy regardless of the bit depth. The information is used as motion compensation calculation accuracy selection information. Hereinafter, the former motion compensation processing in which calculation accuracy is adjusted according to the bit depth is referred to as implementation-oriented motion compensation processing, and the latter motion compensation processing in which calculation accuracy is fixed is referred to as accuracy-oriented motion compensation processing. In the present embodiment, when the former implementation-oriented motion compensation process is selected, the motion compensation calculation accuracy selection information is 0, and when the latter accuracy-oriented motion compensation process is selected, the motion compensation calculation accuracy selection information is It shall be 1. However, the combination of the selected motion compensation process and motion compensation calculation accuracy selection information is not limited to these. Further, the determination method of the motion compensation calculation accuracy selection information is not particularly limited, and may be determined prior to the encoding process assuming an application in which the present encoding device and the corresponding decoding device are used. It may be selected by a user (not shown). For example, when it is assumed that the encoding apparatus of the present embodiment is used in an application in which calculation accuracy is important, the motion compensation calculation accuracy selection information is set to 1, and otherwise it is set to 0.

  Next, the motion compensation calculation accuracy information generation unit 323 generates motion compensation calculation accuracy information based on the above-described motion compensation calculation accuracy selection information and the bit depth information input from the input unit 102. When the motion compensation calculation accuracy selection information is 0, the difference value between the bit depth of the image and the reference bit depth of 8 bits is used as motion compensation calculation accuracy information. In this embodiment, since the bit depth of the image is 10 bits, the motion compensation calculation accuracy information is 2. When the motion compensation calculation accuracy selection information is 1, the motion compensation calculation accuracy information is 0. However, the combination of the value and meaning of the motion compensation calculation accuracy information is not limited to the above, and when the bit depth of the image is larger than the reference bit depth, it can be shown that the calculation accuracy of the motion compensation processing is improved. It only needs to be information.

  The generated motion compensation calculation accuracy selection information is output to the header encoding unit 304, and the motion compensation calculation accuracy information is output to the prediction unit 305 and the image reproduction unit 308.

  The header encoding unit 304 encodes information necessary for decoding, including the bit depth information input from the input unit 102 and the motion compensation calculation accuracy selection information input from the motion compensation calculation accuracy information generation unit 323, and includes a header. Generate code data. This header encoded data corresponds to the header portion of the bit stream. The generated header code data is output to the integrated encoding unit 111.

  On the other hand, the prediction unit 305 receives the image data in units of blocks divided by the input unit 102 and the motion compensation calculation accuracy information generated by the motion compensation calculation accuracy information generation unit 323. Next, prediction in units of blocks is performed, and prediction information indicating a prediction method such as intra prediction that is intra-frame prediction or inter prediction that is inter-frame prediction is generated. The generation method of the prediction information is not particularly limited, and may be determined based on the similarity between the encoded pixel stored in the frame memory 109 and the pixel of the encoding target block. It may be determined using. The generated prediction method is output to the image reproduction unit 308 and the block encoding unit 110. Based on the generated prediction information, the encoded image stored in the frame memory 109 is appropriately referred to generate a predicted image. When the prediction target image is generated, if the encoding target block is inter prediction encoded, a motion compensation process based on the motion compensation calculation accuracy information is performed. Specifically, one of the calculation formulas used for the motion compensation processing of the sub-pixels in the motion compensation of the color difference signal shown by the above-described formula (1) is the following formula (2) in the present embodiment. become.

_{ab 0,0 = (- 2 × B} -1,0 + 58 × B 0,0 + 10 × B 1,0 -2 × B 2,0) >> shift ... (2)
(However, shift = motion compensation calculation accuracy information, and “>>” represents a bit shift to the right.)
Also in the above equation (2), like the above equation (1), B _{i, j} is an intermediate value for calculating the color difference pixel at the integer pixel position, and ab _0,0 is the color difference pixel at the decimal pixel position. Represents. In Expression (2), the bit shift process to the right by shift is based on motion compensation calculation accuracy information. Therefore, when the implementation-oriented motion compensation process is selected by the motion compensation calculation accuracy information generation unit 323, the expression (2) includes a bit shift process to the right depending on the bit depth, and thus the intermediate value ab _0,0 The range of values that can be taken is constant regardless of the bit depth of the image. On the other hand, when accuracy-oriented motion compensation is selected by the motion compensation calculation accuracy information generation unit 323, the shift value is always 0 in Equation (2), the right bit shift processing is not executed, and the calculation accuracy is maintained. Can be processed.

  Finally, the prediction unit 305 generates a prediction error as a difference between the input image in units of blocks and the generated prediction image, and outputs the prediction error to the transform quantization unit 306.

  The transform quantization unit 306 performs orthogonal transform on the prediction error input from the prediction unit 305, generates transform coefficients, further quantizes the transform coefficients, and generates quantized coefficients. The generated quantization coefficient is output to the inverse quantization inverse transform unit 307 and the block coding unit 110.

  The inverse quantization inverse transform unit 307 dequantizes the quantization coefficient input from the transform quantization unit 306 to reproduce the transform coefficient, and further performs inverse orthogonal transform to the reconstructed transform coefficient to reproduce a prediction error. The reproduced prediction error is output to the image reproduction unit 308.

  The image reproduction unit 308 reproduces a predicted image by appropriately referring to the frame memory 109 based on the prediction information input from the prediction unit 305 and the motion compensation calculation accuracy information input from the motion compensation calculation accuracy information generation unit 323. . When the prediction target image is generated, if the encoding target block is inter prediction encoded, a motion compensation process based on the motion compensation calculation accuracy information is performed. Specifically, similarly to the prediction unit 305, motion compensation processing represented by Expression (2) is performed.

  Then, the image reproduction unit 308 generates a reproduction image from the generated prediction image and the prediction error input from the inverse quantization inverse transformation unit 307. The generated reproduced image is output to the frame memory 109 and stored.

  FIG. 18A shows an example of the bit stream generated in this embodiment. The motion compensation calculation accuracy selection information is included as a motion compensation calculation accuracy selection information code in any of the headers of sequences, pictures, and the like. Similarly, bit depth information is also included in one of the headers as a bit depth information code.

  However, the configuration of the bit stream is not limited to this, and instead of encoding the motion compensation calculation accuracy selection information code as shown in FIG. 18B, a corresponding profile is determined, and the determined profile is used as profile information. You may encode as a code | symbol. For example, it is assumed that there are a main 10-bit profile and a main 10-bit high-precision profile, which correspond to motion compensation calculation accuracy selection information = 0 and motion compensation calculation accuracy selection information = 1, respectively. That is, the implementation-oriented motion compensation process is selected in the main 10-bit profile, and the accuracy-oriented motion compensation process is selected in the main 10-bit high-accuracy profile. In this case, when the motion compensation calculation accuracy selection information is 0, a code indicating a main 10-bit profile may be encoded as a profile information code. Absent.

  FIG. 7 is a flowchart illustrating an encoding process in the image encoding device according to the third embodiment. The same numbers are assigned to portions that perform the same functions as those in FIG. 5 of the first embodiment, and descriptions thereof are omitted.

  In step S722, the motion compensation calculation accuracy information generation unit 323 generates motion compensation calculation accuracy selection information for selecting motion compensation calculation accuracy information indicating the calculation accuracy in the motion compensation process. In this embodiment, the motion compensation calculation accuracy selection information is 0 when the implementation-oriented motion compensation processing is selected, and the motion compensation calculation accuracy selection information is 1 when the latter accuracy-oriented motion compensation processing is selected. Shall be.

  In step S723, the motion compensation calculation accuracy information generation unit 323 generates motion compensation calculation accuracy information based on the motion compensation calculation accuracy selection information generated in step S722 and the bit depth information generated in step S501. In step S704, the header encoding unit 304 encodes information necessary for decoding, including the bit depth information generated in step S501 and the motion compensation calculation accuracy selection information generated in step S502, to generate header code data. Is generated.

  In step S706, the input unit 102 cuts out a square block from the input image data, and the prediction unit 305 performs block unit prediction on the cut out block unit image data. Furthermore, prediction information indicating a prediction method such as intra prediction that is intra-frame prediction or inter prediction that is inter-frame prediction is generated. Based on the generated prediction information, the encoded pixel stored in the frame memory 109 is referred to as appropriate to generate a predicted image. When generating the predicted image, if the target block to be encoded is inter prediction encoded, a motion compensation process based on the motion compensation calculation accuracy information generated in step S723 is performed. Specifically, the motion compensation process represented by the above-described equation (2) is executed. Further, a prediction error is generated as a difference between the input image data in block units and the predicted image.

  In step S707, the transform quantization unit 306 performs orthogonal transform on the prediction error generated in step S706 to generate transform coefficients, further quantizes the generated transform coefficients, and generates quantized coefficients. In step S708, the inverse quantization inverse transform unit 307 performs inverse quantization on the quantization coefficient generated in step S707 to reproduce the transform coefficient, and further performs inverse orthogonal transform to the reconstructed transform coefficient to reproduce a prediction error. .

  In step S709, the image reproduction unit 308 generates a predicted image by appropriately referring to the frame memory 109 based on the prediction information generated in step S706. When generating the predicted image, if the target block to be encoded is inter prediction encoded, a motion compensation process based on the motion compensation calculation accuracy information generated in step S723 is performed. Specifically, similarly to step S706, motion compensation processing represented by equation (2) is performed. Then, a reproduced image is generated from the generated predicted image and the prediction error reproduced in step S 708 and stored in the frame memory 109. In step S711, the image coding apparatus determines whether or not coding of all the blocks in the frame is finished. If finished, the coding process is finished. If not, the next block is selected. The process returns to step S706 as a target.

  With the above configuration and operation, in particular, by encoding the motion compensation calculation accuracy selection information in step S704, a bitstream that can be switched between encoding processes with different calculation accuracy and implementation cost according to the required specifications of the application is generated. be able to.

  In this embodiment, the flow of the encoding process has been described in the order of steps S708, S709, and S510. However, the order is not limited to this, and step S510 may be executed after step S707.

  When the bit depth of the image data to be encoded is 8 bits, the motion compensation calculation accuracy selection information code can be omitted. That is, when the bit depth matches 8 bits, the motion compensation calculation accuracy information is uniquely determined to be 0, and redundant codes can be reduced.

  In the main 10-bit profile and the main 10-bit high-precision profile, the motion compensation calculation accuracy selection information is always 0 in the former, but the latter selects 0/1 by providing a motion compensation calculation accuracy selection information code. A configuration may be used. As a result, the calculation accuracy can be selected even with a high-precision profile.

  In addition, the bitstream generated in the present embodiment is encoded in the order of the motion compensation calculation accuracy selection information code and the bit depth information code as shown in FIG. 18A, but the order is limited to this. Not.

<Embodiment 4>
FIG. 4 is a block diagram showing an image decoding apparatus according to this embodiment. In FIG. 4, the same numbers are assigned to portions that perform the same functions as those in FIG. 2 of the second embodiment, and the description thereof is omitted. In the present embodiment, the decoding of the bitstream generated in the third embodiment will be described as an example.

  A header decoding unit 403 decodes the header code data separated from the bit stream, and reproduces information related to the decoding process. Reference numeral 424 denotes a motion compensation calculation accuracy information setting unit that generates motion compensation calculation accuracy information indicating the calculation accuracy of the motion compensation processing used in the image reproduction unit 407. Reference numeral 406 denotes an inverse quantization inverse transform unit that inversely quantizes the quantization coefficient reproduced by the block decoding unit 205 to reproduce a transform coefficient, and further performs inverse orthogonal transform to reproduce a prediction error. The difference from the inverse quantization inverse transform unit 206 of the second embodiment is that the transform quantization computation accuracy information is not input and the inverse quantization inverse transform process is always performed with the same computation accuracy.

  Reference numeral 407 denotes an image reproduction unit that performs intra prediction, inter prediction, and the like with reference to the frame memory 208 based on the prediction information reproduced by the block decoding unit 205, and generates a prediction error generated by the inverse quantization inverse transformation unit 206. A playback image is generated from The image reproducing unit 207 of the second embodiment is different from the image reproducing unit 207 in that motion compensation calculation accuracy information is input and inter prediction is performed based on the input motion compensation calculation accuracy information.

  An image decoding operation in the image decoding apparatus will be described below. In the present embodiment, the bit stream generated in the third embodiment is decoded.

  The header decoding unit 403 decodes information necessary for decoding from the header code data input from the separation decoding unit 202, and reproduces motion compensation calculation accuracy selection information and bit depth information. The reproduced motion compensation calculation accuracy selection information and bit depth information are output to the motion compensation calculation accuracy information setting unit 424.

  The motion compensation calculation accuracy information setting unit 424 generates motion compensation calculation accuracy information based on the motion compensation calculation accuracy selection information and the bit depth information input from the header decoding unit 403. Similar to the motion compensation calculation accuracy information generation unit 323 of the third embodiment, in this embodiment, when the motion compensation calculation accuracy selection information is 0, the 8-bit difference value that is the bit depth information and the reference bit depth is used. Is motion compensation calculation accuracy information. Since the bit stream generated in the third embodiment is obtained by encoding a 10-bit image, the bit depth information in this embodiment is 10 bits, and the motion compensation calculation accuracy information is 2. On the other hand, when the motion compensation calculation accuracy selection information is 1, 0 is used as the motion compensation calculation accuracy information. However, as in the third embodiment, the combination of motion compensation calculation accuracy selection information and motion compensation calculation accuracy information is not limited to these. The generated motion compensation calculation accuracy information is output to the image reproduction unit 407.

  The inverse quantization inverse transform unit 406 dequantizes the quantized coefficient input from the block decoding unit 205 to reproduce the transform coefficient, and further performs inverse orthogonal transform to the reconstructed transform coefficient to reproduce the prediction error. The reproduced prediction error is output to the image reproduction unit 407.

  The image reproduction unit 407 generates a predicted image by appropriately referring to the frame memory 208 based on the prediction information input from the block decoding unit 205 and the motion compensation calculation accuracy information input from the motion compensation calculation accuracy information setting unit 424. . When generating a predicted image, when the decoding target block is inter-predictively encoded, motion compensation processing based on motion compensation calculation accuracy information is performed. Specifically, the motion cheek award process represented by the above formula (2) is performed. A reproduced image is generated from the generated predicted image and the prediction error input from the inverse quantization and inverse transform unit 406. The generated reproduced image is output to the frame memory 208 and stored.

  FIG. 8 is a flowchart illustrating a decoding process in the image decoding apparatus according to the fourth embodiment. The same numbers are assigned to portions that perform the same functions as those in FIG. 6 of the second embodiment, and descriptions thereof are omitted.

  In step S802, the header decoding unit 403 decodes information necessary for decoding from the header code data separated in step S601, and reproduces motion compensation calculation accuracy selection information and bit depth information. In step S823, the motion compensation calculation accuracy information setting unit 424 generates motion compensation calculation accuracy information based on the motion compensation calculation accuracy selection information and the bit depth information reproduced in step S802. In step S824, the image reproduction unit 407 determines the calculation accuracy in the subsequent motion compensation processing based on the motion compensation transform quantization calculation accuracy information generated in step S823. In step S806, the inverse quantization inverse transform unit 406 dequantizes the quantization coefficient generated in step S605 to reproduce the transform coefficient, and further performs inverse orthogonal transform to the reconstructed transform coefficient to reproduce the prediction error. .

  In step S807, the image reproduction unit 407 generates a predicted image by appropriately referring to the frame memory 208 based on the prediction information reproduced in step S605. When generating the predicted image, if the decoding target block is inter-predictively encoded, a motion compensation process based on the motion compensation information determined in step S824 is performed. Specifically, a motion compensation process represented by the above-described equation (2) is performed. Then, reproduced image data is generated from the generated predicted image and the predicted image reproduced in step S 806 and stored in the frame memory 208. At the same time, the reproduced image data is output from the terminal 209.

  With the above configuration and operation, in particular, by decoding the motion compensation calculation accuracy selection information in step S802, bits that can be decoded with different calculation accuracy and implementation cost according to the required specifications of the application generated in the third embodiment. The stream can be decoded.

  In the present embodiment, the input bit stream is obtained by independently encoding the motion compensation calculation accuracy selection information shown in FIG. 18A, but the present invention is not limited to this. For example, as shown in FIG. 18B, instead of encoding a motion compensation calculation accuracy selection information code, a profile information code indicating a corresponding profile may be encoded. In this case, the motion compensation calculation accuracy information setting unit 424 generates motion compensation calculation accuracy information from the profile information code and the bit depth information.

<Embodiment 5>
FIG. 9 is a block diagram showing an image encoding apparatus according to this embodiment. In FIG. 9, the same numbers are assigned to portions that perform the same functions as those in FIG. 1 of the first embodiment and FIG. 3 of the third embodiment, and description thereof is omitted.

  Reference numeral 902 denotes an input unit that analyzes the bit depth of input image data and divides it into square blocks. Unlike the input unit 102 of the first embodiment, the bit depth information is also output to the motion compensation calculation accuracy information generation unit 323. Reference numeral 904 denotes a header encoding unit that encodes information necessary for decoding a bitstream including bit depth information of an image to generate header code data.

  An image encoding operation in the image encoding apparatus will be described below.

  The input unit 902 analyzes the bit depth of the input image data, and outputs the bit depth information to the subsequent motion compensation calculation accuracy information generation unit 323, the transform quantization calculation accuracy information generation unit 103, and the header encoding unit 904. However, the bit depth information may be separately provided from the outside and input to the motion compensation calculation accuracy information generation unit 323, the transform quantization calculation accuracy information generation unit 103, and the header encoding unit 904, respectively. Further, the input image data is divided into square blocks and output to the prediction unit 305.

  The header encoding unit 904 first selects bit depth information from the input unit 902, motion compensation calculation accuracy selection information from the motion compensation calculation accuracy information generation unit 323, and transform quantization calculation accuracy selection from the transform quantization calculation accuracy information generation unit. Enter information. Then, information necessary for decoding such as the input information is encoded to generate header code data. This header encoded data corresponds to the header part of the bit stream and is output to the integrated encoding unit 111.

  FIG. 19A shows an example of the bit stream generated in the present embodiment. The transform quantization operation accuracy selection information is included as a transform quantization operation accuracy selection information code, and the motion compensation operation accuracy selection information is included as a motion compensation operation accuracy selection information code in one of the headers of a sequence, a picture, and the like. Similarly, bit depth information is also included in one of the headers as a bit depth information code.

  However, the configuration of the bitstream is not limited to this, and a corresponding profile is determined instead of encoding the motion compensation calculation accuracy selection information code and the transform quantization calculation accuracy selection information code as shown in FIG. However, it may be encoded as a profile information code. For example, there are a main 10-bit profile and a main 10-bit high accuracy profile, and the main 10-bit profile corresponds to transform quantization calculation accuracy selection information = 0 and motion compensation calculation accuracy selection information = 0. The main 10-bit high accuracy profile corresponds to transform quantization calculation accuracy selection information = 1 and motion compensation calculation accuracy selection information = 1. That is, the implementation-oriented conversion quantization process and the implementation-oriented motion compensation process are selected in the main 10-bit profile, and the precision-oriented conversion quantization process and the accuracy-oriented motion compensation process are selected in the main 10-bit high-accuracy profile. In this case, when the transform quantization calculation accuracy selection information and the motion compensation calculation accuracy selection information are 0, a code indicating the main 10-bit profile is encoded as a profile information code. Moreover, when the transform quantization calculation accuracy selection information and the motion compensation calculation accuracy selection information are 1, a code indicating a main 10-bit high accuracy profile may be encoded as a profile information code.

  FIG. 13 is a flowchart illustrating an encoding process in the image encoding device according to the third embodiment. The parts having the same functions as those of FIG. 5 of the first embodiment and FIG. 7 of the third embodiment are given the same numbers, and the description thereof is omitted.

  In step S1301, the input unit 902 analyzes the bit depth of the input image data and generates bit depth information. In step S1304, the header encoding unit 904 encodes information necessary for decoding to generate header code data. The information necessary for decoding includes the bit depth information generated in step S1301, the transform quantization calculation accuracy selection information generated in step S502, and the motion compensation calculation accuracy selection information generated in step S722.

  With the above configuration and operation, information on the calculation accuracy of the encoding processing generated in steps S502 and S722 is encoded in step S1304. As a result, it is possible to generate a bitstream that can be switched between encoding processes with different calculation accuracy and implementation cost according to the required specifications of the application.

  In this embodiment, the flow of the encoding process has been described in the order of steps S508, S709, and S510. However, the order is not limited to this, and step S510 may be executed after step S507.

  Further, when the bit depth of the image data to be encoded is 8 bits, the transform quantization calculation accuracy selection information code and the motion compensation calculation accuracy selection information code can be omitted. That is, when the bit depth is 8 bits, the transform quantization calculation accuracy information and the motion compensation calculation accuracy information are uniquely determined to be 0, and redundant codes can be reduced.

  In addition, in the main 10-bit high-accuracy profile described above, a configuration in which a transform quantization calculation accuracy selection information code and a motion compensation selection information code are provided to select 0/1 may be used. This makes it possible to select calculation accuracy even with a high-precision profile.

  In this embodiment, in step S508 in FIG. 13, the calculation accuracy in the inverse quantization / inverse transform process is determined based on the transform quantization calculation accuracy information generated in step S503. However, the calculation accuracy obtained in step S507 is determined. Of course you can use.

  Further, as shown in FIG. 19A, the bit stream generated in the present embodiment is encoded in the order of transform quantization calculation accuracy selection information code, motion compensation calculation accuracy selection information code, and bit depth information code. However, the order is not limited to this.

  Furthermore, in this embodiment, the motion compensation calculation accuracy information generation unit 323 and the transform quantization calculation accuracy information generation unit 103 are provided independently, but as shown in FIG. 11, only one calculation accuracy information generation unit 1143 is provided. It is also possible to take In this case, the calculation accuracy information generated by the calculation accuracy information generation unit 1143 is input to the prediction unit 1105, the transform quantization unit 1106, the inverse quantization inverse transform unit 1107, and the image reproduction unit 1108, and the input calculation accuracy is input. Processing based on the information is performed. Also, the header encoding unit 1104 encodes calculation accuracy selection information and bit depth information.

  In this case, a flowchart showing the corresponding encoding process is as shown in FIG. In FIG. 15, step S1501 analyzes the bit depth by the input unit 1101 as in step S501, and step S1542 generates calculation accuracy selection information for transform quantization processing and motion compensation processing. In step S1543, calculation accuracy information of the transform quantization process and the motion compensation process is generated. Step S1504 encodes calculation accuracy selection information. In step S1506 and step S1509, motion compensation processing is performed based on the calculation accuracy information generated in step S1543. In step S1507, transform quantization processing is performed based on the calculation accuracy information generated in step S1543. In step S1508, inverse quantization inverse transformation processing is performed based on the calculation accuracy information generated in step S1543.

  Thus, independent transform quantization calculation accuracy information and motion compensation calculation accuracy information are used for transform quantization and motion compensation, but common calculation accuracy information can be used.

  In that case, an example of the generated bit stream is shown in FIG. The calculation accuracy selection information is included as a calculation accuracy selection information code in one of the headers of a sequence, a picture, and the like, and the bit depth information is also included in one of the headers as a bit depth information code. Further, similarly to the above-described embodiment, instead of encoding the calculation accuracy selection information as the calculation accuracy selection information code, a corresponding profile may be determined, and the determined profile may be encoded as profile information. In that case, an example of the generated bitstream is shown in FIG.

<Embodiment 6>
FIG. 10 is a block diagram showing an image encoding apparatus according to this embodiment. In FIG. 10, the same numbers are assigned to portions that perform the same functions as those in FIG. 2 of the embodiment and FIG. 4 of the embodiment 4, and the description thereof is omitted.

  Reference numeral 1003 denotes a header decoding unit that decodes header code data separated from the bit stream and reproduces information related to the decoding process.

  An image decoding operation in the image decoding apparatus will be described below. In the present embodiment, the bit stream generated in the fifth embodiment is decoded.

  The header decoding unit 1003 decodes information necessary for decoding from the header code data input from the separation decoding unit 202, and reproduces motion compensation calculation accuracy selection information, transform quantization calculation accuracy selection information, and bit depth information. The reproduced motion compensation calculation accuracy selection information is output to the motion compensation calculation accuracy information setting unit 424, and the reproduced transform quantization operation accuracy selection information is output to the transform quantization operation accuracy information setting unit 204. The reproduced bit depth information is output to the transform quantization calculation accuracy information setting unit 204 and the motion compensation calculation accuracy information setting unit 424.

  FIG. 14 is a flowchart illustrating a decoding process in the image decoding apparatus according to the sixth embodiment. The parts having the same functions as those in FIG. 6 of the second embodiment and FIG. 8 of the fourth embodiment are given the same numbers, and the description thereof is omitted.

  In step S1402, the header decoding unit 1003 decodes information necessary for decoding from the header code data separated in step S601, and reproduces motion compensation calculation accuracy selection information, transform quantization calculation accuracy selection information, and bit depth information. .

  With the above configuration and operation, it is possible to decode the motion compensation calculation accuracy selection information and the transform quantization calculation accuracy selection information regarding the decoding process, particularly in step S1402. As a result, it is possible to decode a bitstream that can be decoded with different calculation accuracy and implementation cost in accordance with the required specifications of the application generated in the fifth embodiment.

  In the present embodiment, the input bitstream is obtained by independently encoding the transform quantization calculation accuracy selection information and the motion compensation calculation accuracy selection information shown in FIG. 19A. However, the present invention is not limited to this. For example, as shown in FIG. 19B, instead of encoding the transform quantization calculation accuracy selection information code and the motion compensation calculation accuracy selection information code, a profile information code indicating a corresponding profile is encoded. It may be. In this case, the motion compensation computation accuracy information setting unit 424 and the transform quantization computation accuracy information setting unit 204 generate motion compensation computation accuracy information and transform quantization computation accuracy information from the profile information code and the bit depth information.

  Furthermore, in this embodiment, the transform quantization calculation accuracy information setting unit 204 and the motion compensation calculation accuracy information setting unit 424 are provided independently, but as shown in FIG. 12, only one calculation accuracy information setting unit 1244 is provided. It is also possible to take In this case, the calculation accuracy information set by the calculation accuracy information setting unit 1244 is input to the inverse quantization inverse transform unit 1206 and the image reproduction unit 1207, and processing based on the input calculation accuracy information is performed.

  In this case, a flowchart showing the corresponding decoding process is as shown in FIG. In FIG. 16, S1602 decodes calculation accuracy selection information. Step S1643 reproduces the calculation accuracy information of the inverse quantization inverse transform process and the motion compensation process. In step S1644, the calculation accuracy information of the inverse quantization inverse transform process and the motion compensation process is reproduced and determined based on the reproduced calculation accuracy information. In steps S1606 and S1607, inverse quantization inverse transformation processing and motion compensation processing are performed based on the respective calculation accuracy information determined in step S1644. Thus, independent transform quantization calculation accuracy information and motion compensation calculation accuracy information are used for transform quantization and motion compensation, but common calculation accuracy information can be used.

  In that case, an example of the input bit stream is shown in FIG. 19C, and the transform quantization calculation accuracy selection information and the motion compensation calculation accuracy selection information are encoded independently. However, the present invention is not limited to this. Similarly to the above-described embodiment, instead of the calculation accuracy selection information encoded as the calculation accuracy selection information code, the corresponding profile may be encoded as profile information. In that case, an example of the input bit stream is shown in FIG.

<Embodiment 7>
1, 2, 3, 4, 9, 10, 11, and 12 have been described in the above embodiment as being configured by hardware. However, the processing performed by each processing unit shown in these drawings may be configured by a computer program.

  FIG. 20 is a block diagram illustrating a configuration example of computer hardware applicable to the image display device according to each of the above embodiments.

  The CPU 2001 controls the entire computer using computer programs and data stored in the RAM 2002 and the ROM 2003, and executes the processes described above as performed by the image processing apparatus according to each of the above embodiments. That is, the CPU 2001 functions as each processing unit shown in FIGS. 1, 2, 3, 4, 9, 10, 11, and 12.

  The RAM 2002 has an area for temporarily storing computer programs and data loaded from the external storage device 2006, data acquired from the outside via the I / F (interface) 2007, and the like. Further, the RAM 2002 has a work area used when the CPU 2001 executes various processes. That is, for example, the RAM 2002 can be allocated as a frame memory or can provide various other areas as appropriate.

  The ROM 2003 stores setting data of the computer, a boot program, and the like. The operation unit 2004 includes a keyboard, a mouse, and the like, and various instructions can be input to the CPU 2001 by a user of the computer. The output unit 2005 displays the processing result by the CPU 2001. The output unit 2005 is configured by a liquid crystal display, for example.

  The external storage device 2006 is a large-capacity information storage device represented by a hard disk drive device. In the external storage device 2006, an OS (operating system) and functions of each unit shown in FIGS. 1, 2, 3, 4, 9, 10, 11, and 12 are realized by the CPU 2001. A computer program is stored. Furthermore, each image data as a processing target may be stored in the external storage device 2006.

  Computer programs and data stored in the external storage device 2006 are appropriately loaded into the RAM 2002 under the control of the CPU 2001 and are processed by the CPU 2001. The I / F 2007 can be connected to a network such as a LAN or the Internet, and other devices such as a projection device and a display device. The computer can acquire and send various information via the I / F 2007. Can be. Reference numeral 2008 denotes a bus connecting the above-described units.

  The operation having the above-described configuration is controlled by the CPU 2001 centered on the operation described in the above flowchart.

  The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, etc.) of the system or apparatus reads the program. It is a process to be executed.

Claims (23)

An image encoding device for encoding an image, comprising:
Deriving means for deriving a coefficient by performing conversion processing on the difference between the target pixel and the predicted pixel, data corresponding to the coefficient, and the minimum value and the maximum value related to the coefficient are determined or fixed by the bit depth A flag indicating whether or not, encoding means for encoding,
An image encoding apparatus comprising: The image coding apparatus according to claim 1, wherein the transform process is an orthogonal transform process. The image coding apparatus according to claim 1, wherein the deriving unit derives the coefficient for each block included in the image. The image encoding apparatus according to claim 1, further comprising: a prediction processing unit that performs prediction processing based on the target pixel to derive the prediction pixel. The image encoding apparatus according to claim 1, wherein the data corresponding to the coefficient is data indicating the coefficient that has been subjected to quantization processing. The said flag is a flag which shows whether the minimum value and the maximum value which the said coefficient can take are determined by the bit depth, or are fixed. The one of the Claims 1-5 characterized by the above-mentioned. Image encoding device. The said flag is a flag which shows whether the minimum value and the maximum value which can be taken in the said conversion process are determined by a bit depth, or are fixed. The one of Claims 1-6 characterized by the above-mentioned. Image coding apparatus. The image coding apparatus according to any one of claims 1 to 7, wherein the bit depth is a bit depth of the image to be encoded. The image encoding apparatus according to claim 1, wherein the encoding unit encodes information indicating the bit depth. An image decoding device for decoding an image from a bitstream,
Decoding means for decoding from the bitstream data corresponding to coefficients related to pixels of the image and a flag indicating whether the minimum and maximum values related to the coefficients are determined by a bit depth or fixed ,
An image decoding apparatus comprising: a derivation unit that performs conversion processing and derives a difference between a target pixel and a predicted pixel based on the coefficient. The image decoding apparatus according to claim 10, wherein the transform process is an inverse orthogonal transform process. The image decoding device according to claim 10 or 11, wherein the derivation unit derives the difference based on the coefficient for each block included in the image. The image decoding apparatus according to claim 10, further comprising a generation unit configured to generate the image from the prediction pixel and the difference. The coefficient is a transform coefficient after the inverse quantization process is performed,
The derivation unit derives the difference by performing the transformation process on the transformation coefficient after the inverse quantization process is performed. The difference according to any one of claims 10 to 13, The image decoding device described. 15. The flag according to claim 10, wherein the flag is a flag indicating whether a minimum value and a maximum value that can be taken by the coefficient are determined by a bit depth or are fixed. Image decoding device. The said flag is a flag which shows whether the minimum value and the maximum value which can be taken in the said conversion process are determined by a bit depth, or are fixed. The one of Claims 10-15 characterized by the above-mentioned. Image decoding apparatus. The image decoding device according to any one of claims 10 to 16, wherein the bit depth is a bit depth of the image to be decoded. The image decoding device according to any one of claims 10 to 17, wherein the information indicating the bit depth is decoded from the bit stream. The said derivation | leading-out means performs the said conversion process according to the minimum value and maximum value according to the said flag, and derives | leads-out the said difference from the said coefficient. The any one of Claims 10-18 characterized by the above-mentioned. Image decoding apparatus. An image encoding method for encoding an image, comprising:
A derivation step of deriving a coefficient by performing a conversion process on the difference between the target pixel and the predicted pixel, data corresponding to the coefficient, and the minimum and maximum values related to the coefficient are determined or fixed by the bit depth An encoding step for encoding a flag indicating whether or not
An image encoding method characterized by comprising: An image decoding method for decoding an image from a bitstream,
A decoding step of decoding from the bitstream data corresponding to coefficients related to pixels of the image and a flag indicating whether the minimum and maximum values related to the coefficients are determined by a bit depth or fixed; ,
An image decoding method comprising: a derivation step of performing a conversion process and deriving a difference between the target pixel and the predicted pixel based on the coefficient.   A program that causes a computer to function as each unit of the image encoding device according to any one of claims 1 to 9.   A program that causes a computer to function as each unit of the image decoding device according to any one of claims 10 to 19.

Images