JP2008153802A - Moving picture encoding device and moving picture encoding program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that a conventional moving picture encoding device determines a quantization value without considering encoding types of adjacent MBs, so that block distortion between adjacent MBs differing in encoding type is generated and then picture quality deteriorates. <P>SOLUTION: In a motion picture encoding device, a quantization determining unit 112 inputs intra/inter MB types for each MB from an ME/MC type determination unit 111, determines a quantization value of an MB input from a peripheral MB information storage unit 113 by using peripheral MB information on the MB, and makes an adjustment in accordance with an average quantization value input from a code amount control unit 114. The peripheral MB information storage unit 113 holds the quantization value and MB type (intra/inter) of the MB determined by a quantization determining unit 112. Furthermore, the peripheral MB information storage unit 113 outputs the peripheral MB information to the quantization determining unit 112 as needed. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a moving image coding apparatus and a moving image coding program, and in particular, each screen of a moving image signal is divided into blocks each having a predetermined number of pixels as a unit, and the correlation between pictures in units of the blocks. The present invention relates to a moving picture coding apparatus and a moving picture coding program for generating moving picture coded data by performing orthogonal transformation, quantization, and variable length coding in units of blocks.

  Video encoding technology for moving images is becoming more and more important because video signals with a large amount of information can be transmitted at high speed using a transmission medium with limited bandwidth, or can be recorded for a long time on a recording medium with limited recording capacity. It has become. For example, MPEG (Moving Picture Experts Group) -1, MPEG-2, MPEG-4, H.261, H.263, and H.264 are known as compression coding techniques for moving images. It has been. In these moving image compression coding technologies, each frame of a moving image signal is divided in units of blocks, which are image areas of a predetermined number of pixels, and motion compensated interframe prediction using correlation between pictures in units of blocks. The amount of information is reduced by processing such as the above, and moving image coded data is generated by performing orthogonal transform such as DCT (Discrete Cosine Transform), quantization, and variable length coding in units of blocks.

  In a moving image encoding apparatus that generates such moving image encoded data, it is desired to efficiently increase the encoding efficiency and improve the image quality of the encoded moving image data. Conventionally, a moving picture coding apparatus having means for reflecting a degree (activity) in a change amount of a quantization parameter has been proposed (see, for example, Patent Document 1).

  FIG. 4 is a block diagram showing the overall configuration of the moving picture encoding apparatus described in Patent Document 1. In FIG. 4, the moving image signal is converted into a digital image signal by the A / D conversion circuit 11, and the screen rearrangement circuit 12 sequentially encodes each frame image signal in accordance with the GOP (Group of Picture) structure. The data are rearranged and supplied to the arithmetic circuit 13, the motion prediction / compensation circuit 22, and the activity calculation circuit 23. When the input frame image signal is inter-coded, the arithmetic circuit 13 generates image data indicating a difference between the input frame image signal and the predicted frame image data from the motion prediction / compensation circuit 22. When the input frame image signal is intra-coded, the input frame image signal is supplied to the orthogonal transformation circuit 14 as image data as it is.

  The orthogonal transform circuit 14 supplies a signal obtained by orthogonal transform of the input image data in units of 4 × 4 pixel blocks to the quantization circuit 15, and here, according to the quantization scale from the rate control circuit 25. Quantization is performed. The lossless encoding circuit 16 variable-length-encodes the quantized image data output from the quantization circuit 15 and stores it in the buffer 17, and also encodes the motion vector MV from the motion prediction / compensation circuit 22 or the difference thereof. And store it in the header data.

  The quantized image data output from the quantization circuit 15 is inversely quantized by the inverse quantization circuit 18, is subjected to filter processing for removing block distortion by the deblocking filter 19, and is further subjected to the inverse orthogonal transform circuit 20. Then, the inverse orthogonal transform is performed in units of 4 × 4 pixel blocks, decoded into image data, and stored in the frame memory 21. The motion prediction / compensation circuit 22 generates a motion vector MV and reference image data based on the image data from the frame memory 21 and the image data from the screen rearrangement circuit 12, and the motion vector MV is a lossless encoding circuit. 16, the reference image data is output to the arithmetic circuit 13, respectively.

  On the other hand, the activity calculation circuit 23 calculates the image complexity (activity) of the input moving image signal based on the image data from the screen rearrangement circuit 12 based on a predetermined method, and calculates the calculated activity Nact as the ΔQ calculation circuit 24. Here, the variation data ΔQ of the quantization parameter is calculated according to a predetermined arithmetic expression according to Nact. The rate control circuit 25 adds the change amount data ΔQ and the reference data from the buffer 17 to generate the quantization parameter QP.

  Next, a method for determining the quantization value set using the activity calculation circuit 23, the ΔQ calculation circuit 14, and the rate control circuit 25 will be described with reference to the flowchart of FIG. In FIG. 5, step S501 divides input moving image data into units called GOPs composed of a plurality of pictures, and indicates the start point of a loop in units of GOPs. In step S502, the GOP code amount distribution is determined from the set bit rate. Step S503 indicates the start point of a loop in units of pictures in the GOP.

  In step S504, code amount distribution per picture in the GOP is performed. In Patent Document 1, weighting is performed according to the type of picture (I picture, P picture, B picture), and the code amount per picture is determined. Step S505 indicates the start point of a unit loop called a macroblock (MB) made up of 16 × 16 pixels in the picture.

  Next, in step S506, processing for determining code amount distribution in MB units is performed. In Patent Document 1, the code amount allocated to a picture is simply divided by the total number of MBs in the picture. Here, the quantization value is determined based on the determined code amount. Subsequently, in step S507, a quantization value determination process reflecting the activity is performed. In the process of step S507, first, the activity calculation circuit 23 calculates the complexity (activity) of the image using the luminance value of the input moving image data, and calculates the activity Nact of each MB from the tendency of the entire picture. calculate. The value of this activity Nact is an index for measuring the complexity of the image in the picture, and a large activity value means that the image is complicated. In patent document 1, when an image is complicated, even if a quantized value is set to be coarser than other portions, it is difficult to recognize image degradation with the naked eye, and the quantized value is corrected using activity Nact. It is carried out.

  Subsequently, when encoding is performed using the quantized value determined in step S507 in step S508, the generated code amount is excessive or insufficient as compared with the code amount given to the MB in advance. Execute the checkout process. In Patent Document 1, settlement is performed using the following equation.

Next MB code amount = (PicRate−encoded MB code amount) ÷ number of remaining MBs However, in the above equation, “PicRate” is a code amount to be allocated per picture. That is, since the code amount of a picture is determined, MBs encoded later in the picture are affected by excess or deficiency. In Patent Document 1, since the MB encoding order is encoded from the upper left to the lower right of the picture, the quantization value in MB units is affected by the previous encoding as it goes to the lower right. Become.

  Step S509 indicates the end of the MB unit loop in the picture. For all MBs in the picture, the processes in steps S506 to S508 are repeated for each MB. Subsequently, in step S510, as in step S508, the excess or deficiency of the code amount in units of pictures is settled. Step S511 is the end of a loop for each picture in the GOP. The processes in steps S504 to S510 are performed for each picture in the GOP. In the subsequent step S512, the excess or deficiency of the code amount in GOP units is settled as in step S510. Step S513 is the end of the loop in GOP units.

  In the conventional moving picture coding apparatus described in Patent Document 1 that performs the above operation, the image characteristic is ignored, and the quantization value is determined in order from the upper left to the lower right, and the complexity of the image is shown. By introducing an activity and using it as an index when determining a quantized value, uniform image quality within a picture can be obtained.

JP 2004-194076 A

  Here, the difference in image quality when encoding is performed using equivalent quantization values in intra encoding and inter encoding will be described with reference to FIG. FIG. 6A shows a simple processing flow of intra coding (intra compression), and FIG. 6B shows a simple processing flow of inter coding (inter compression).

  As shown in FIG. 6A, in intra coding (intra compression), each process of transformation 602, quantization 603, inverse quantization 604, and inverse transformation 605 is sequentially performed on the input current image 601. The decoded image 606 can be obtained. That is, in the case of intra compression, it can be said that the amount of information to be reduced varies depending on the magnitude of the quantization value, and the generated decoded image 606 changes greatly.

  On the other hand, as shown in FIG. 6B, inter coding (inter compression) is created from MC by performing motion compensation prediction (MC) 610 using a reference picture for the current image 601. The compression processing is performed separately for the MC image 611 and a residual component 612 that is a difference between the current image 601 and the MC image 611. The MC image 611 can be created using a motion vector value from a reference picture, and is not actually compressed. That is, in the inter compression, only the residual component 612 obtained by the MC 610 is subjected to the transformation 613, quantization 614, inverse quantization 615, and inverse transformation 616 sequentially, and the residual component after the inverse transformation and the MC image are processed. 611 and a decoded image 617 can be obtained.

  In this inter compression, since the MC image 611 exists even when the residual component becomes 0 by the quantization 614, the decoded image 617 is compensated. Therefore, it can be said that the decoded image generated by the magnitude of the quantized value has less influence than the intra compression.

  As described above, when intra compression and inter compression are performed with the same quantized value, if the image quality of the inter compression reference picture is somewhat clean, the decoded image of the inter compression is compared with the image of the intra compression. Is beautiful. That is, in the case of compressing with an equivalent quantized value, a difference in image quality occurs between intra compression and inter compression.

  However, in the conventional video encoding device described in Patent Document 1, an activity that is an image characteristic of a picture is calculated, and the activity is reflected when the quantization value is determined. When trying to determine the quantized value, the Nth MB is affected by the quantized values of the MBs determined up to the (N-1) th. For example, consider a case where the encoding type of the (N-1) th MB is an inter MB and the encoding type of the Nth MB is an intra MB. It is assumed that when the quantization value of the (N-1) th MB is determined, the code amount is equal to the scheduled code amount.

  Then, the quantized value of the next MB (Nth) becomes the same as the (N−1) th quantized value, which causes a block distortion as described above. That is, in the conventional video encoding device described in Patent Document 1, since the quantization value is determined without considering the encoding type of the adjacent MB, there is a problem that the image quality may be deteriorated. .

  The present invention has been made in view of the above points, and it is an object of the present invention to provide a moving picture coding apparatus and a moving picture coding program that can improve coding efficiency more efficiently and improve image quality.

In order to achieve the above object, the first invention divides one screen of a moving image signal into block units, which are image regions having a predetermined number of pixels, and decodes the encoded moving image signal and encoded data. The difference signal between the prediction signal obtained by motion compensation prediction using the reference image signal and the moving image signal to be encoded is sequentially subjected to orthogonal transform, quantization, and variable length coding in units of blocks. Then, in the moving picture coding apparatus for generating the coded data of each screen that is inter-coded,
When all the blocks of one screen to be encoded are composed of a predetermined first block and a plurality of second blocks other than the first block, the first quantized value candidates of the first block are Whether the block is a first coding type that performs inter coding using information of a corresponding block of a reference image signal or a second coding type that performs intra coding without using a reference image signal First coding type information to be stored in advance, and a plurality of second block quantization value candidates and whether the second block is the first coding type or the second coding Information storing means for sequentially storing all the second encoding type information indicating whether the type is a block unit, and a first one of the plurality of second blocks respectively obtained from the information storing means Adjacent to 2 blocks Based on the first or second encoding type information of the side block and the first or second quantization value candidate, and the second encoding type information of the second block acquired from the video signal, 2nd quantization value which determines the 2nd quantization value candidate of 2 blocks, and preserve | saves the 2nd coding type information of 2nd block with the determined 2nd quantization value candidate in an information storage means Quantization value determination means for sequentially performing value candidate determination / storing operation for all second blocks, and a picture average based on the code amount assigned when inter-coding one screen to be encoded A calculation means for calculating a quantized value, and reads out first and second quantized value candidates of all the first and second blocks of one screen to be encoded from the information storage means, and calculates an average value thereof. The calculated average value is the picture average quantization value. Correction means for correcting the first and second quantized value candidates of the first and second blocks read from the information storage means so as to be equivalent to generate quantized values for each block; Based on the quantized value for each block generated by the correcting means, quantization is performed on the orthogonally transformed difference signal in units of blocks.

  In order to achieve the above object, the second invention divides one screen of a moving image signal into block units, which are image regions having a predetermined number of pixels, and uses the moving image signal to be encoded and the encoded data. For the difference signal between the prediction signal obtained by motion compensation prediction using the decoded reference image signal and the moving image signal to be encoded, orthogonal transform, quantization, and variable length encoding are performed in block units. In a moving image encoding program that is sequentially performed and generates encoded data of each screen that has been inter-encoded, the computer causes each of the means of the first invention to function.

  In the first and second inventions, the encoding type information of the block to be encoded, the encoding type information of neighboring blocks adjacent to the block, and the quantization value candidate are acquired, and the block to be encoded is acquired. After determining the quantization value candidates for one picture of the encoding target screen, the average quantization value candidate of all the blocks of the encoding target screen is equal to the average quantization value of the picture The quantized value candidate for each block is corrected so that the quantized value for each block is generated.

  According to the present invention, when a quantization value is determined in a predetermined block unit, information on blocks around the block to be encoded (encoding type information and quantization value candidates) is used to quantize the block unit. After obtaining the quantized value candidates for all blocks in one screen, the average value of the quantized value candidates of all blocks in the picture to be encoded on one screen is corrected to be equal to the average quantized value of that picture. Since the quantization value is generated, the block distortion between the blocks when the encoding type of the block to be encoded and the neighboring block adjacent to each other is different can be reduced, and the subjective image quality can be improved. realizable.

  Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a moving picture encoding apparatus according to the present invention. In this embodiment, a frame memory 101 that accumulates one or more frames of moving image signals to be encoded, a frame output unit 102, a subtractor 103 that obtains a difference signal, an orthogonal transformer 104 that performs orthogonal transformation, A quantizer 105 that quantizes an input signal according to a quantization value, an encoding circuit 106 that performs variable-length encoding processing, and an inverse quantizer 107 that performs inverse quantization on the quantized image data. An inverse orthogonal transformer 108 that performs inverse orthogonal transformation on the image data after inverse quantization, an adder 109, a frame memory 110, a motion estimation (ME) process for calculating a motion vector in MB units, motion A ME / MC / type determination unit 111 that performs compensation (MC) processing and type determination processing of whether the processing target MB is an intra MB or an inter MB, and further includes a code amount control unit 114 and a quantization determination unit. 12 and a peripheral MB information storage unit 113.

  The frame memory 101 holds several pieces (several frames) of image data obtained by converting a moving image signal to be encoded into a digital signal, and outputs the held image data as a frame when necessary. The data is output to the subtractor 103 and the ME / MC / type determination unit 111 via the unit 102. The subtractor 103 generates difference data between the image data output from the frame output unit 102 and the image data subjected to motion compensation prediction processing output from the ME / MC / type determination unit 111 by subtraction, and the generated difference Data is output to the orthogonal transformer 104.

  The orthogonal transformer 104 performs, for example, DCT on the image data output from the frame output unit 102 or the difference data output from the subtractor 103 as an example of orthogonal transformation, and converts the data into frequency components (DCT coefficient information). To do. Data obtained by orthogonal transformation by the orthogonal transformer 104 is input to the quantizer 105. In this embodiment, DCT is used as orthogonal transform. However, other orthogonal transform methods that facilitate encoding of input data may be used.

  The quantizer 105 quantizes the data input from the orthogonal transformer 104 with the quantization value input from the quantization determination unit 112. The quantized data output from the quantizer 105 is supplied to the encoding circuit 106 and the inverse quantizer 107, respectively. The encoding circuit 106 performs a known variable length encoding (VLC) process on the quantized data input from the quantizer 105. In the present embodiment, the encoding circuit 106 performs variable length encoding using the VLC table, and outputs the data obtained by encoding to the outside as encoded image data. Also, the encoding circuit 106 outputs parameters used in the code amount control unit 114 to the code amount control unit 114. In this embodiment, the encoding circuit 106 uses a VLC table, but other variable length encoding methods may be used.

  The inverse quantizer 107 performs inverse quantization on the quantized data input from the quantizer 105 by performing an operation opposite to that performed by the quantizer 105. The data obtained by inverse quantization by the inverse quantizer 107 is supplied to the inverse orthogonal transformer 108, where an operation reverse to the operation performed by the orthogonal transformer 104 is performed and inverse orthogonal transformation is performed. If the image data extracted by inverse orthogonal transform by the inverse orthogonal transformer 108 is image data to be intra-encoded, it is supplied to the frame memory 110, and if it is image data to be inter-encoded, an adder 109 is supplied. To be supplied.

  The adder 109 adds the image data after the inverse orthogonal transform output from the inverse orthogonal transformer 108 and the data output from the ME / MC / type determination unit 111 to generate decoded image data. The generated decoded image data is output to the frame memory 110. The frame memory 110 holds decoded image data input from the inverse orthogonal transformer 108 or the adder 109. The frame memory 110 outputs the stored image data as reference image data to the ME / MC / type determination unit 111 via the frame output unit 102 as necessary.

  The ME / MC / type determination unit 111 acquires the decoded image data held in the frame memory 110 as reference image data, and acquires the image data to be encoded from the frame memory 101 via the frame output unit 102. Then, the motion of the image data to be encoded is estimated in MB units, a motion vector indicating the relative positional relationship between the reference image data and the corresponding MB of the image data to be encoded is calculated (ME), and motion compensation is performed based on the motion vector (MC) is performed to create picture prediction image data.

  The ME / MC / type determination unit 111 also determines whether the MB of the encoding target image data input from the frame output unit 102 is an intra encoding target intra MB or an inter encoding target inter MB. The determination (type determination) is also performed. This determination is made by comparing the code amount when the MB of the image data to be encoded is inter-encoded by the frame output unit 102 and the code amount when the intra-encoding is performed, and the determination result of the type with the smaller code amount Get.

  The motion vector calculated by the ME / MC / type determination unit 111 is output to the encoding circuit 106, the picture prediction image data created by the motion compensation is output to the subtracter 103 and the adder 109, and the type determination result is the quantum Is output to the conversion determination unit 112.

  The code amount control unit 114 uses the parameters input from the encoding circuit 106, determines the initial quantized value of the first MB of the picture to be encoded, and the average quantized value Qave in the picture. The initial quantization value and the average quantization value Qave are output to the quantization determination unit 112. Note that any control may be performed as long as the code amount control is such that the average quantization value is calculated in units of pictures and the quantization value of the leading MB is determined.

  The quantization determination unit 112 receives the MB type of intra / inter in MB units from the ME / MC / type determination unit 111, uses the peripheral MB information of the corresponding MB input from the peripheral MB information storage unit 113, MB quantization value candidates are determined and adjusted so as to follow the average quantization value input from the code amount control unit 114. This will be described in detail later.

  The peripheral MB information storage unit 113 holds at least the MB quantization value candidate and the MB type (intra / inter) determined by the quantization determination unit 112 for each MB. Also, the peripheral MB information storage unit 113 outputs the peripheral MB information to the quantization determination unit 112 as necessary.

  Next, the operation of the present embodiment will be described with reference to the flowchart of FIG. A detailed description of each step is the same as the above description of the block diagram of FIG. First, image data obtained by converting a moving image signal to be encoded into a digital signal is input and stored in the frame memory 101 (step S201). The subsequent step S202 indicates the starting point of the loop for one picture unit, and the processing of step S203 to step S216, which will be described later, up to step S217 indicating the end of the picture unit loop is performed for each picture to be encoded. Step S203 subsequent to step S202 indicates the start point of the MB-unit loop in the picture, and the processing in steps S204 to S208 described later up to step S209 that indicates the end of the MB-unit loop in the picture is to be encoded. This is performed in units of MB for all MBs in the picture.

  Subsequently, the frame output unit 102 determines whether the encoding target picture output from the frame memory 101 is an intra picture or an inter picture (step S204). If it is an intra picture, the process proceeds to step S208. If it is an inter picture, the process proceeds to step S205.

  When the picture to be encoded is an inter picture, the ME / MC / type determination unit 111 encodes reference image data input from the frame memory 110 and encoding input from the frame memory 101 via the frame output unit 102. Using the target image data, the motion of the image data to be encoded is estimated in MB units, and a motion vector indicating the relative positional relationship between the reference image data and the corresponding MB of the image data to be encoded is calculated. (ME), motion compensation (MC) is performed on the basis thereof to generate picture prediction image data, and the MB of the image data input from the frame memory 101 via the frame output unit 102 is the intra MB to be intra-encoded. Or the type of inter MB to be inter-coded (step S2). 5).

  Subsequently, by the quantization determination unit 112, the intra / inter MB type input in units of MB from the ME / MC / type determination unit 111, and the peripheral MB information of the corresponding MB input from the peripheral MB information storage unit 113, Then, using the initial quantized value input from the code amount control 111, the quantized value candidate of the corresponding MB is determined as will be described later (step S206), and the determined quantized value candidate and type information are determined. Is stored in the peripheral MB information storage unit 113 (step S207). On the other hand, if it is determined in step S204 that the picture to be encoded is an intra picture, the quantization determination unit 112 determines a quantization value in MB units based on the output of the code amount control unit 111 (step S204). S208).

  When the processing of steps S204 to S208 is performed in units of MB for all MBs in the encoding target picture, whether or not the encoding target picture is an inter picture is determined based on the output of the frame output unit 102. It judges again (step S210), and when it is an intra picture, it moves to step S212, and when it is an inter picture, it moves to step S211.

  In step S211, correction processing for making the quantized value candidate of each MB in the inter picture to be coded determined in step S206 equal to the picture average quantized value input from the code amount control unit 114 To finally determine the quantization value in MB of the inter picture to be encoded. Subsequently, after the orthogonal transformation is performed on the image data or the difference data input to the orthogonal transformer 104, the quantization based on the quantization value in MB determined by the quantizer 105 in step S208 or S211 is performed by MB. Image compression that is performed in units and includes variable length coding by the coding circuit 106 for the quantized data, inverse quantization by the inverse quantizer 107, and inverse orthogonal transform by the inverse orthogonal transformer 108 for the inverse quantized data. Processing is performed (step S212).

  Subsequently, it is determined whether or not the image data of the picture to be encoded, which is extracted by inverse orthogonal transform by the inverse orthogonal transformer 108, is a reference picture used at the time of inter coding (step S213), and is a non-reference picture. If so, the process proceeds to step S216. If it is a reference picture, the process proceeds to step S214.

  Step S214 is a step of creating a decoded image. If the picture is a completely intra-coded picture, no processing is performed. If the picture is a reference picture used for inter coding, the adder 109 performs inverse orthogonality. The inverse orthogonal transform image data (decoded error image data) output from the converter 108 and the data generated by motion compensation output from the ME / MC / type determination unit 111 are added and decoded. Image data is created (step S214). The generated decoded image data is output and stored in the frame memory 110 (step S215).

  Then, the code amount control unit 114 uses the parameters input from the encoding circuit 106, and determines the initial quantization value of the first MB of the picture to be encoded next and the average quantization value in the picture. A unit code amount control process is performed (step S216). The determined initial quantization value and the average quantization value are output to the quantization determination unit 112. Here, the average quantization value (Qave described later) in the above picture is obtained by encoding each picture in the GOP from the code amount allocated in the GOP until immediately before encoding the encoding target picture. Based on the code amount obtained by subtracting the obtained total code amount and the number of remaining pictures before encoding in the GOP, the code amount calculated by a predetermined algorithm is assigned to the encoding target picture. Is.

  In this embodiment, code amount control in units of MB is not performed. However, code amount control in units of MB is performed after steps S207 and S208, and the quantization value of the next MB is calculated. Quantized values may be used.

  Next, the quantization determination unit 112 will be described in detail. The quantization determination unit 112 receives the intra / inter type determination result of the encoding target MB from the ME / MC / type determination unit 111, and from the neighboring MB information storage unit 113, as shown in FIG. The quantized value candidate and type information of the upper MB 302 located at the upper adjacent position to the current encoding target MB 301 and the quantized value candidate and type information of the left MB 303 located at the left adjacent position are input. The code amount control unit 114 receives the initial quantization value of the first MB and the average quantization value Qave corresponding to the picture type of the inter picture, and uses these input data to determine the current encoding target. MB301 quantization value candidates are determined, and the determined quantization value candidates are stored in the peripheral MB information storage unit 113 together with the picture type. Since 1 MB at the left end and the upper end of one screen does not have a peripheral MB, the peripheral MB information storage unit 113 stores the picture average quantized value as a quantized value candidate of the MB.

  As soon as quantization value candidates for all MBs in the encoding target picture are determined and stored in the peripheral MB information storage unit 113, the quantization of all MBs of the encoding target picture is performed from the peripheral MB information storage unit 113. The value candidates are read and the average value is calculated, and the quantized value candidates of each MB are corrected as a whole until the calculated average value is equal to the average quantized value Qave of the encoding target picture. Generate quantized values.

  For example, it is assumed that quantization value candidates for all MBs in the encoding target picture are determined using peripheral information. In this case, the average quantization value candidate of all MBs in the encoding target picture is set to “20”. The quantization value to be set is smaller by “2” (= 22−20) when the predetermined average quantization value Qave of the input encoding target picture is “22”. Therefore, in this case, the quantization value candidates of all MBs in the encoding target picture are increased by “2”, and the quantization value candidates of all MBs in the picture are corrected to be equal to the average quantization value Qave. As a result, the quantization value of each MB is obtained.

  A table of quantization determination methods for the target MB is shown in Table 1.

なお、表１において、カレントＭＢは現在の符号化対象のＭＢであり、イントラ補正値（ＩｎｔｒａＱ）とは、ある量子化値をイントラ量子化値に換算する数値であり、インター補正値（ＩｎｔｅｒＱ）とは、ある量子化値をインター量子化値に換算する数値である。

In Table 1, the current MB is the current encoding target MB, and the intra correction value (IntraQ) is a numerical value for converting a certain quantized value into an intra quantized value, and an inter correction value (InterQ). Is a numerical value for converting a certain quantized value into an inter quantized value.

  In the present embodiment, the intra correction value (IntraQ) and the inter correction value (InterQ) are “1 / 1.1” when the P picture is encoded, InterQ) is “1.1”, and when the B picture is encoded, the intra correction value (IntraQ) is “1 / 1.4” and the inter correction value (InterQ) is “1.4”.

  As described above, according to the present embodiment, when the quantization determination unit 112 determines the quantization value candidates of each MB in the inter picture in units of MB, the neighboring MBs ( After obtaining quantization value candidates for all MBs in units of MB using the information on each MB adjacent to the left side and the upper side, the average value of the quantization value candidates for all MBs in the picture to be encoded in one screen is Since the quantized value candidate is corrected in each MB unit so as to be equal to the average quantized value Qave of the picture and the quantized value is generated in each MB unit, block distortion generated between adjacent MBs is reduced. The subjective image quality can be improved.

  In this embodiment, the intra correction value and the inter correction value are fixed values, but may be arbitrarily changed depending on image characteristics. If there is no neighboring MB such as a picture boundary, only the existing MB is used to calculate a quantization value candidate. In this embodiment, the upper MB and the left MB are set as the peripheral MBs. However, the quantized value candidates are calculated using the upper left MB and the upper right MB, which have already been calculated, as peripheral information. May be. In this embodiment, the MB type is used to determine the quantized value candidate. However, the quantized value candidate is determined by weighting using information such as the motion vector value of the neighboring MB and the type determination result. May be.

  The present invention includes a moving picture coding program for realizing each component other than the frame memories 101 and 109 of the moving picture coding apparatus shown in FIG. 1 by a computer. In this case, the moving image encoding program may be taken into the computer from the recording medium, or may be delivered via the communication network and taken into the computer.

It is a block diagram of one embodiment of a moving picture coding apparatus of the present invention. It is a flowchart explaining operation | movement of this invention apparatus. It is explanatory drawing of an example of peripheral MB of the present MB. It is a block diagram of the conventional moving image encoding device described in Patent Document 1. 6 is a code amount control flowchart of the apparatus of FIG. It is a figure explaining the difference between intra compression and inter compression.

Explanation of symbols

101, 110 Frame memory 103 Subtractor 104 Orthogonal transformer 105 Quantizer 106 Coding circuit 107 Inverse quantizer 108 Inverse orthogonal transformer 109 Adder 111 ME / MC / type determining unit 112 Quantization determining unit 113 Peripheral MB information Storage unit 114 Code amount control unit

Claims (2)

It is obtained by dividing one screen of a moving image signal into blocks, which are image areas having a predetermined number of pixels, and performing motion compensation prediction using the moving image signal to be encoded and a reference image signal decoded from the encoded data. The difference signal between the encoded prediction signal and the video signal to be encoded is sequentially subjected to orthogonal transform, quantization, and variable length encoding in units of blocks, and the inter-encoded screens In a video encoding device that generates encoded data,
When all the blocks of one screen to be encoded are composed of a predetermined first block and a plurality of second blocks other than the first block, the first quantization of the first block A value candidate and a first encoding type in which the block performs the inter encoding using information of a corresponding block of the reference image signal, or a second code that performs intra encoding without using the reference image signal First coding type information indicating whether the second block is a quantization type is stored in advance, and the quantized value candidates of the plurality of second blocks, and the second block is the first code. Information storage means for sequentially storing all the second encoding type information indicating whether the encoding type or the second encoding type, in units of blocks;
The first or second encoding type information of the neighboring block adjacent to one second block of the plurality of second blocks and the first or second obtained respectively from the information storage unit Determining the second quantized value candidate of the second block based on the quantized value candidate and the second encoding type information of the second block acquired from the video signal; The second quantization value candidate determination / storage operation for storing the second encoding type information of the second block in the information storage unit together with the determined second quantization value candidate is performed for all the second quantization values. Quantization value determining means for sequentially performing the blocks,
Calculation means for calculating a picture average quantization value based on a code amount assigned when inter-encoding one screen to be encoded;
Read the first and second quantized value candidates of all the first and second blocks of one screen of the encoding target from the information storage unit, calculate the average value, and calculate the average value Quantize each block by correcting the first and second quantized value candidates of the first and second blocks read from the information storage means so that is equal to the picture average quantized value. Correcting means for generating a value, and performing the quantization on the orthogonally transformed difference signal in units of blocks based on the quantized value for each block generated by the correcting means. A video encoding apparatus characterized by the above. It is obtained by dividing one screen of a moving image signal into blocks, which are image areas having a predetermined number of pixels, and performing motion compensation prediction using the moving image signal to be encoded and a reference image signal decoded from the encoded data. The difference signal between the encoded prediction signal and the video signal to be encoded is sequentially subjected to orthogonal transform, quantization, and variable length encoding in units of blocks, and the inter-encoded screens In a moving image encoding program for generating encoded data by a computer,
In the computer,
When all the blocks of one screen to be encoded are composed of a predetermined first block and a plurality of second blocks other than the first block, the first quantization of the first block A value candidate and a first encoding type in which the block performs the inter encoding using information of a corresponding block of the reference image signal, or a second code that performs intra encoding without using the reference image signal First coding type information indicating whether the second block is a quantization type is stored in advance, and the quantized value candidates of the plurality of second blocks, and the second block is the first code. Each of the plurality of second blocks obtained from the information storage means for sequentially storing all the second encoding type information indicating whether the encoding type or the second encoding type is in block units. The first in The first or second encoding type information and the first or second quantization value candidate of the neighboring block adjacent to the block, and the second block of the second block acquired from the moving image signal The second quantization value candidate of the second block is determined based on the coding type information, and the second encoding of the second block is determined together with the determined second quantization value candidate. A first step of sequentially performing a second quantized value candidate determining / storing operation for storing type information in the information storing unit for all the second blocks;
The second quantized value candidate determined in the first step is stored in the information storage unit, and the second encoding type information of the second block to be encoded is stored in the information storage unit. A second step;
A third step of calculating a picture average quantization value based on a code amount assigned when inter-encoding one screen to be encoded;
A fourth step of reading out the first and second quantized value candidates of all the first and second blocks of one screen of the encoding target from the information storage unit and calculating an average value thereof;
The first and second quantized values of the first and second blocks read from the information storage means so that the average value calculated in the fourth step is equivalent to the picture average quantized value. A fifth step of correcting candidates and generating a quantized value for each block;
And causing the computer to execute the quantization on the block-by-block basis for the orthogonally transformed difference signal based on the quantization value for each block generated by the fifth step. A video encoding program.

Images