JP2006129067A - Moving image encoding device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce mosquito noise that occurs around a moving person. <P>SOLUTION: A quantization coefficient control circuit of an MPEG-2 encoder to which this invention is applied performs coefficient control so as to make a quantization coefficient small when activity of a macroblock of a quantization target is low and to make the quantization coefficient large when the activity is high on the basis of steps 1, 2 and 3 of a TM (test model) 5. Further, the quantization coefficient control circuit compares the activity of the macroblock of the quantization target with activity of macroblocks around the macroblock of the quantization target, determines whether the macroblock of the quantization target is a macroblock located at a boundary between an area with low activity and an area with high activity, and makes a quantization coefficient value small when the macroblock is determined to be a macroblock at the boundary part. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a moving picture coding apparatus and method for coding a moving picture signal by controlling a quantization coefficient in units of macroblocks.

  In MPEG-2, TM5 (Test Model 5; ISO / IEC JTC / SC29 / WG11 (1993)), a quantization coefficient control method for improving image quality is proposed. Specifically, TM5 has a hierarchy (step 1) for optimal code distribution to each picture, a hierarchy (step 2) for feedback control of quantization coefficients based on the capacity of the virtual buffer, and a visual characteristic. And a hierarchy (step 3) for controlling the quantization coefficient in units of macroblocks.

  Here, in step 3, the image is more finely quantized with a flat image portion that is visually prominently deteriorated (a portion with low activity), and is more coarsely quantized with a complex image portion (a portion with high activity) that is less prone to deterioration. Thus, the quantization coefficient obtained in step 2 is changed. Specifically, in step 3, the complexity of the pattern is indexed by activity, and the quantization coefficient obtained in step 2 is changed.

Japanese Patent Laid-Open No. 10-66067

  By the way, in step 3 of TM5, the macroblock located at the boundary between the flat part and the complicated part of the pattern tends to be judged as having a complicated pattern, and is roughly quantized. Therefore, for example, on a screen on which a person who is doing sports, for example, is displayed on a white wall, that is, a screen on which an object image with a large movement is displayed on a flat background image, the background image and the object Mosquito noise was likely to occur at the border with the image.

  The present invention solves the above-described problems, and an object of the present invention is to provide a moving image encoding apparatus and method in which image quality deterioration due to quantization is reduced.

  In the moving picture coding apparatus and method according to the present invention, the activity of the macroblock to be quantized is compared with the activity of the surrounding macroblocks, and the quantization coefficient for the macroblock to be quantized is calculated based on the comparison result. To do.

  Furthermore, in the moving picture coding apparatus and method according to the present invention, a predetermined arithmetic expression is set such that the quantization coefficient is decreased when the activity of the macroblock to be quantized is low, and the quantization coefficient is increased when the activity is high. Based on the above, the quantization coefficient is controlled for each macroblock, and the activity of the macroblock to be quantized is compared with the activity of the surrounding macroblock, and the macro of the boundary portion between the low activity area and the high activity area is compared. It is determined whether or not the block is a block, and if it is determined that the block is a macro block at the boundary portion, the value of the quantization coefficient calculated based on the predetermined arithmetic expression is reduced.

  With the moving picture coding apparatus and method according to the present invention as described above, it is possible to reduce image quality degradation due to quantization.

  Furthermore, in the moving picture coding apparatus and method according to the present invention, the quantization is fine in the flat part of the image, coarsely in the complex part of the pattern, and fine in the boundary part of the moving object image. Quantization can be efficiently performed according to the above and generation of mosquito noise can be reduced.

  An MPEG-2 (ISO / IEC13818-2 Information Technology-Generic Coding of moving pictures and associated audio information: Video) video encoder to which the present invention is applied will be described below.

(Overall block configuration)
FIG. 1 is a configuration diagram of a video encoder 10 to which the present invention is applied.

  As shown in FIG. 1, the video encoder 10 receives spatial area image data (pixel data (Pixel)), performs an encoding process corresponding to the MPEG-2 standard, and outputs an encoded stream (Stream). .

  The video encoder 10 includes a motion estimation unit 11 that calculates a motion vector, a subtraction unit 12, a discrete cosine transform (DCT) unit 13, a quantization unit 14, and a VLC unit 15 that performs variable length coding (VLC) processing. A quantization coefficient control unit 16 that controls the quantization coefficient, an inverse quantization unit 17, an inverse discrete cosine transform (IDCT) unit 18, an addition unit 19, a memory 20, and a motion compensation unit 21. ing.

  Pixel data (Pixel) is input to the motion estimation unit 11 in units of macroblocks, and motion vectors for each macroblock are calculated. When calculating the motion vector, the motion estimation unit 11 calculates the activity and ME residual of the macroblock, and calculates the motion vector using these.

  The activity is an index representing the complexity of the pattern, and is represented based on the variance. That is, a macroblock with a large activity has a large variation in the level of each pixel, and a macroblock with a small activity has a small variation in the level of each pixel and is flattened. A specific arithmetic expression for calculating the activity will be described later.

  The ME residual is an index indicating the total level remaining after motion estimation. The ME residual is defined as, for example, an absolute value sum or a square value sum of a difference value between an input image and a reference image after a motion compensation process using a motion vector at the time of compression encoding. When there is no correlation between the input image and the reference image, the ME residual is large, and when there is a correlation between the input image and the reference image, the ME residual is small.

  Image data (pixel) is input to the subtraction unit 12 in units of macroblocks. In addition, when the encoding process using the correlation between frames is performed on the macroblock (pixel data (pixel)) to be encoded, the subtraction unit 12 also receives the prediction block from the motion compensation unit 21. Entered. The subtraction unit 12 subtracts the prediction block from the input image data when performing processing on the inter macroblock. The subtraction unit 12 outputs the input image data as it is when processing the intra macroblock.

  The DCT unit 13 performs discrete cosine transform on the macroblock (image data (pixel)) output from the subtracting unit 12 to generate DCT coefficient data (coef) that is image data in the frequency domain. The DCT unit 13 outputs the generated coefficient data (coef) to the quantization unit 14.

  The quantization unit 14 performs a quantization process on the input coefficient data (coef) using the quantization coefficient Q. That is, the quantization unit 14 divides the input coefficient data (coef) by the quantization coefficient Q to obtain a quotient, and outputs the quotient. The coefficient data (coef) is irreversibly compressed by being quantized by the quantization unit 14. If the quantization coefficient Q is large, the amount of code to be output is small and the compression rate is high. Conversely, if the quantization coefficient Q is small, the amount of code to be output is large and the compression rate is low.

  The VLC unit 15 reads the coefficient data (coef) in units of macroblocks quantized by the quantization unit 14 by a predetermined scanning method, and performs encoding by run length encoding. At the same time, the VLC unit 15 performs fixed-length coding or the like on the motion vector and other system information. The VLC unit 15 multiplexes these data according to the MPEG-2 system and generates a coded stream (Stream) composed of one bit string. The VLC unit 15 outputs the generated encoded stream (Stream) to the outside.

  The quantization coefficient control unit 16 calculates a quantization coefficient Q used in the quantization process of the quantization unit 14 and supplies the quantization coefficient Q to the quantization unit 14. The quantization coefficient control unit 16 determines the quantization coefficient Q by referring to the bit rate of the encoded stream (stream) output from the VLC unit 15, the activity calculated by the motion estimation unit 11, and the ME residual. . Detailed processing contents of the quantization coefficient control unit 16 will be described later.

  Of the coefficient data output from the quantization unit 14, the inverse quantization unit 17 receives coefficient data of a macroblock of a frame that can be a reference image for interframe prediction. The inverse quantization unit 17 performs inverse quantization processing on the input data using the quantization coefficient used when the data is quantized.

  The IDCT unit 18 performs inverse discrete cosine transform on the coefficient data (coef) output from the inverse quantization unit 17 to generate image data (pixel) in the spatial domain.

  The image data (pixel) output from the IDCT unit 18 is input to the adding unit 19. In addition, when the input image data is a macroblock encoded using the correlation between frames, the prediction block is input from the motion compensation unit 21 to the addition unit 19. The adder 19 adds the prediction block to the input image data when performing processing on the inter macroblock. The adder 19 outputs the input image data as it is when processing the intra macroblock. The adder 19 stores the output image data in the memory 20 as a reference image.

  The motion compensation unit 21 performs motion compensation on the reference image stored in the memory 20 using a motion vector, and generates a prediction block. Note that, when an intra macroblock is stored, the motion compensation unit 21 outputs the image data as it is as a prediction block without performing motion compensation. The motion compensation unit 21 supplies the prediction block to the subtraction unit 12. Moreover, the prediction block used as a reference image is also supplied to the addition part 19 among prediction blocks.

  In the MPEG-2 video encoder 10 as described above, when one macroblock is input, the type of the macroblock is first determined, and if it is an intra macroblock, the input macroblock is directly converted into DCT. If it is an inter macroblock, a prediction block is generated, and the prediction block is subtracted from the input macroblock and supplied to the DCT unit 13.

  Subsequently, in the video encoder 10, the DCT unit 13 performs orthogonal transform on the input macroblock unit pixel data by DCT calculation to obtain frequency domain coefficient data, and then the quantization unit 14 performs the quantization coefficient Q. Quantization processing is performed. Data (coefficient data) in units of macroblocks subjected to DCT calculation and quantization is supplied to the VLC unit 15.

  In the video encoder 10, the VLC unit 15 performs variable-length coding of the coefficient data, multiplexes the coded stream, generates a coded stream, and outputs it.

(Quantization coefficient controller)
Next, processing contents of the quantization coefficient control unit 16 will be described.

  The quantization coefficient control unit 16 is a unit that calculates a quantization coefficient Q used in the quantization process of the quantization unit 14 and supplies the quantization coefficient Q to the quantization unit 14. In the video encoder 10, the bit rate of the encoded stream to be output is controlled by the quantization coefficient Q. That is, when the quantization coefficient Q is increased, the bit rate is controlled to decrease, and when the quantization coefficient Q is decreased, the bit rate is controlled to increase.

--About TM5--
The quantization coefficient control unit 16 calculates the quantization coefficient Q by using Step 1, Step 2 and Step 3 of TM5.

  First, Step 1, Step 2 and Step 3 of TM5 will be described.

  The following two points can be mentioned as the characteristics of MPEG-2 quantization.

  (1) In MPEG-2, the amount of generated code differs greatly for each picture type (I, P, B). In addition, the rate of change of the generated code amount according to the change of the quantization coefficient Q is also greatly different for each picture type.

  (2) The quantization noise of the I picture and the P picture propagates to the P picture and the B picture that refer to this, but the quantization noise of the B picture does not propagate to other pictures. Therefore, making the quantization coefficient Q of all pictures uniform does not necessarily optimize the overall image quality.

  In MPEG-5 TM5, the above two features are used, so that the optimum bit allocation, rate control, and adaptive quantization can be performed while reducing image quality degradation, so that Step 1 → Step 2 → Step 3 It is specified that quantization coefficients are calculated hierarchically in order.

  In Step 1 of TM5, bit allocation to each picture is performed. Specifically, in step 1, the allocated bit amount for each picture in the GOP is allocated based on the bit amount R allocated to a picture that has not yet been encoded in the GOP including the picture to be allocated. To do.

In step 2 of TM5, three types of virtual buffers (VBV buffers) set independently for each picture type are used in order to match the actual generated code amount with the target bit amount for each picture obtained in step 1. Based on the capacity of the VBV buffer, the quantization scale code Q _{j is obtained} by macroblock unit feedback control.

In step 3 of TM5, the quantization obtained in step 2 is performed so that the quantization is finer at the flat part where deterioration is visually noticeable and the deterioration is more coarse at the complicated part of the pattern where deterioration is not visually noticeable. The scale code Q _j is changed according to the activity of the macroblock, and the quantized scale code mquant _j considering the visual characteristics is calculated.

--About Step 3 of TM5--
The process of step 3 of TM5 will be further described.

  For the calculation of the activity of a certain macroblock j, four blocks of the luminance component in the frame DCT coding mode of the macroblock j and four blocks of the luminance component in the field DCT coding mode are used.

First, in step 3 of TM5, the variance Var _sblk is obtained for each of the above eight blocks as shown in the equation (11).

P (bar) in Equation (11) is an average value of each block (8 × 8 = 64 pixels) as shown in Equation (12). P _k is a pixel value of each pixel.

  Subsequently, in step 3, as shown in equation (13), the minimum value is selected from the variances of the eight blocks, and 1 is added to the selected minimum value.

The value obtained by Expression (13) is the activity act _j of the macroblock j. In Expression (13), min () is an expression for selecting the minimum value. The reason why the minimum value is selected is that if there is a portion with high flatness even in only a portion of the macroblock, quantization is performed finely.

When the activity act _j is obtained, then in step 3 of TM5, a normalized activity Nact _j having a value of 0.5 or more and 2.0 or less is calculated based on the arithmetic expression shown in the equation (14). To do.

  In the equation (14), avg_act is an average activity of the picture encoded immediately before. That is, in the equation (14), if the activity actj is the same as the average activity avg_act of the previous picture, it is normalized so as to be 1.

Finally, in step 3, as shown in equation (15), the quantized scale code Q _j obtained in step 2 is multiplied by the normalized activity Nact _j to obtain the final quantized scale code for the macroblock j. Determine mquant _j .

In TM5, the coefficient data after DCT is quantized with the quantization scale code mquant _j obtained in step 3 above.

As a result, the deterioration is visually noticeable as shown in the area A of FIG. 2, that is, the quantization coefficient is reduced in the macroblock (flat portion) with a small activity, and the deterioration is visually shown in the area B of FIG. The quantization scale code obtained in step 2 is changed so that the quantization coefficient is increased in macroblocks (complex parts of the pattern) that have large activity that is not easily noticeable. The quantization scale code mquant _j can be calculated.

--Specific quantization flow--
Next, a specific processing procedure of the quantization coefficient control unit 16 will be described with reference to the flowchart of FIG.

  When the first macroblock in the picture (the macroblock on the upper left of the screen) is input, the quantization coefficient control unit 16 starts processing from the following step S11.

  First, in step S11, the quantization coefficient control unit 16 assigns a code amount to the picture based on step 1 of TM5.

Subsequently, in step S12, the quantization coefficient control unit 16 calculates a quantization scale code Q _j for a macroblock to be quantized (hereinafter referred to as a target macroblock) based on step 2 of TM5. Thereafter, based on Step 3 of TM5, a quantization scale code mquant _{j in} which visual characteristics for the target macroblock are taken into consideration is calculated.

  Subsequently, in step S13, the quantization coefficient control unit 16 determines whether or not the target macroblock is the uppermost or leftmost macroblock of the screen. That is, the quantization coefficient control unit 16 is a macroblock in the uppermost row of the screen as shown in FIG. 4C or a macroblock in the leftmost column of the screen as shown in D of FIG. Determine whether.

  If the target macroblock is the top or left macroblock of the screen (YES in step S13), the quantization coefficient control unit 16 proceeds to step S15 without performing step S14 (dynamic object boundary processing), and the target macroblock If the block is not the top or left macroblock of the screen (NO in step S13), the process proceeds to step S14 (dynamic object boundary processing).

  Subsequently, in step S14, the quantization coefficient control unit 16 performs dynamic object boundary processing. In the boundary processing of dynamic objects, it is determined whether or not the target macroblock is a macroblock located at the boundary between a moving object image and a flat background image. Processing for correcting the quantization coefficient obtained in step 3 of TM5 to be small is performed. Details of the boundary processing of the dynamic object will be described later.

  Subsequently, in step S15, the quantization coefficient control unit 16 determines whether or not the target macroblock is the last macroblock of the picture (that is, the lower right macroblock of the screen).

  If the target macroblock is the last macroblock of the picture, the quantization coefficient control unit 16 ends the flow. If the target macroblock is not the last macroblock, the quantization coefficient control unit 16 returns to step 12 and performs the step for the next macroblock. The process of S12-S15 is performed.

  Next, the dynamic object boundary processing in step S14 will be described in more detail.

  In the boundary processing of the dynamic object in step S14, an object image (hereinafter referred to as a dynamic object) having a large movement such as a person playing sports, for example, as shown in E of FIG. And so on. If it is a boundary macroblock, the quantization coefficient obtained in step 3 of TM5 is determined. This is a process of adjusting so as to make it smaller.

  FIG. 6 shows a more detailed flowchart of the boundary processing of the dynamic object in step S14.

First, in step S21, the quantization coefficient control unit 16 determines whether or not the normalized activity Nact _j obtained in the process of step 3 of TM5 is smaller than 1. The reason for determining whether or not the normalization activity Nact _j is smaller than 1 is that if it is smaller than 1, since the quantization coefficient has already been reduced in the process up to step 3 of TM5, the quantization coefficient is newly reduced. This is because such correction is unnecessary.

If the normalized activity Nact _j is smaller than 1, the process proceeds to step S31. If the normalized activity Nact _j is not smaller than 1, the process proceeds to the following step S22.

  Subsequently, in step S22, the quantization coefficient control unit 16 compares the activity of the target macroblock with macroblocks around the target macroblock (hereinafter referred to as a surrounding macroblock), and the target macroblock Then, a determination is made as to whether or not the flat background image is located at the boundary portion between the pattern and the complex object image.

  As a comparison judgment method, for example, the activity of the surrounding macroblock is subtracted from the activity of the target macroblock, and if the subtraction result is equal to or greater than a predetermined threshold value, the target macroblock is determined as a flat background image and a picture. Is located at the boundary with the complex object image.

  Note that the surrounding macroblocks may be all surrounding macroblocks in the eight directions of the target macroblock (up / down / left / right, upper left, upper right, lower left, lower right), and each direction may be compared and determined. In general, since the quantization process starts from the macro block at the upper left of the screen and is sequentially scanned by the raster scan in the row direction, as shown in FIG. Only the macro block of is set as the surrounding macro block. That is, in this example, the activity of the macroblock on the left is subtracted from the activity of the target macroblock, and the subtraction result is equal to or greater than the predetermined threshold, or the activity of the macroblock on the upper side from the activity of the target macroblock. Is subtracted, and it is determined whether the subtraction result is equal to or greater than a predetermined threshold.

  Of the four luminance blocks (8 × 8 pixels) included in one macroblock, the activity used to detect the boundary portion uses the largest value among the four luminance blocks for comparison determination. ing. The reason for this is that TM5 uses the minimum activity of the eight blocks in the target macroblock and saves the image quality by detecting whether there is only a flat part in the macroblock. Yes. On the other hand, if the maximum activity of the surrounding macroblocks and the maximum activity of the target macroblock are used, it can be detected that only a part of the target macroblock includes a complicated picture. That is, in step S21, when the maximum activity of the surrounding macroblocks is subtracted from the maximum activity of the target macroblock, and the subtraction result is equal to or greater than a predetermined threshold, the target macroblock is determined between the flat background image and the object image. Judge as the boundary.

  In addition, when the target macroblock is the uppermost row or the leftmost column of the screen in the above-described step S12, the boundary processing of the dynamic object is not performed, but this is the encoding process in this step S21. This is because the activities of the surrounding macroblocks cannot be referred to in the macroblocks in the first row (upper end of the screen) and the first column (left end of the screen).

  If the quantization coefficient control unit 16 determines in step S22 that the target macroblock is not located at the boundary between the flat background image and the object image having a complicated pattern, the process proceeds to step S31. On the other hand, if it is determined that the target macroblock is located at the boundary between the flat background image and the object image having a complicated pattern, the process proceeds to the following step S23.

  Subsequently, in step S23, the quantization coefficient control unit 16 determines whether or not the currently encoded picture is an I picture. If it is not an I picture, that is, if it is a P picture or a B picture, the process proceeds to step S24, and if it is an I picture, the process proceeds to step S33.

  Subsequently, in step S24, the quantization coefficient control unit 16 determines whether the object image included in the target macroblock is a dynamic image with a large motion or a static image with a small motion. To do. That is, it is determined that the target macroblock is already positioned at the boundary between the flat background image and the object image with a complicated pattern in step S22. However, the object image with the complicated pattern further moves. It is determined whether the image is an object image.

  As a specific determination method, for example, it is determined whether or not the ME residual of the target macroblock is greater than or equal to a predetermined threshold value. At the boundary between the moving object and the flat background image, the motion vector tends to be off and the ME residual tends to increase. Conversely, if there is no movement even at the boundary between the flat background image and the object image with a complicated pattern, the ME residual becomes small. Since mosquito noise frequently occurs at the boundary between a moving image and a flat partial image, this determination is made here.

  Note that the ME residual is not calculated for the I picture. Therefore, in step S23 in the previous stage, a P picture and a B picture are selected, and the motion determination process is performed only in the case of the P picture and the B picture.

  If the quantization coefficient control unit 16 determines in step S21 that the target macroblock is not located at the boundary portion of the object image with large motion, the process proceeds to step S31. On the other hand, if it is determined that the target macroblock is located at the boundary portion of the object image having a large movement, the process proceeds to step S32.

In step S31, a process of determining that the quantization scale code mquant _j calculated based on step 3 of TM5 is used as it is as the quantization coefficient Q is performed.

That is, when it is determined that the normalized activity Nact _j is smaller than 1 (YES in step S21), when it is determined that the target macroblock is not located at the boundary portion between the flat background image and the object image having a complicated pattern ( If NO in step 22), and if it is determined that the movement of the object having a complicated pattern is not large (NO in step 24), the quantization scale code mquant _j obtained in step 3 of TM5 is directly used as the quantization coefficient. Used as Q. That is, the process for reducing the quantization coefficient is not performed.

If the normalized activity Nact _j is smaller than 1, it is determined that the macroblock is located on the flat portion, and the quantization coefficient obtained in step 3 of TM5 is used as it is.

  In step S32, it is determined that the quantization scale code Qj calculated based on step 2 of TM5 is used as the quantization coefficient Q. That is, in step S32, quantization is performed without using the process of step 3 of TM5.

That is, the normalization activity Nact _j is 1 or more (NO in step S21), and the target macroblock is located at the boundary between the object image having a complicated pattern and the flat background image (step S22). If the object whose pattern is complex is moving greatly (YES in step S24), the value of TM5, which is smaller than the quantization scale code mquant _j obtained in step 3 of TM5, is The quantization scale code Qj calculated based on step 2 is set as a quantization coefficient Q.

  By setting in this way, the quantization coefficient Q becomes a smaller value than in the case of TM5, so that it is difficult to generate mosquito noise.

In step S33, the calculation formula of the normalized activity Nact _j is changed to the following formula (16), and then the quantization scale code mquant _j is obtained based on the above formula (15), and the quantization coefficient Q is converted to this quantum coefficient Q Set to the scaled scale code mquant _j .

  In the normalization activity of Equation (14) used in TM5, Nact takes a range from 0.5 to 2 but in Equation (16), a normalization activity takes a small range of 2/3 to 3/2. It becomes. By taking such a small range, the difference value of the quantization coefficient for each macroblock becomes small, the variation of the quantization coefficient Q becomes small, and it becomes possible to prevent the occurrence of mosquito noise.

  When the process of step S31, step S32, or step S33 is completed, the quantization scale control unit 16 performs the subsequent process of step S34.

  In step S34, the determined quantization coefficient Q is transmitted to the quantization unit 14.

  The quantization scale control unit 16 can reduce the mosquito noise generated around the object image having a complex moving picture by performing the processing as described above. Further, the original effect of Step 3 of TM5 can be maintained except for the boundary portion of the object image having a complicated pattern.

1 is a block configuration diagram of an MPEG-2 video encoder to which the present invention is applied. FIG. It is a figure which shows the image part (A) where deterioration is visually conspicuous, that is, activity is small, and the image part (B) where activity is difficult to visually deteriorate. It is a flowchart which shows the processing content of a quantization coefficient control process. It is a figure which shows the macroblock (C) of the row | line | column of the upper end of a screen, and the macroblock (D) of the column of the left end of a screen. It is a figure which shows the macroblock (E) located in the boundary of a flat background image and an object image with the pattern with a motion complicated. It is a flowchart which shows the processing content of the boundary process of a dynamic object. It is a figure which shows a surrounding macroblock.

Explanation of symbols

10 video encoders, 11 motion estimation units, 12 subtraction units, 13 DCT units, 14 quantization units, 15 VLC units, 16 quantization coefficient control units, 17 inverse quantization units, 18 IDCT units, 19 addition units, 20 memories, 21 Motion compensation unit

Claims (6)

In a video encoding device that divides a screen in a matrix form in units of blocks (macroblocks) composed of a predetermined number of pixels and encodes a video signal in units of the divided macroblocks,
A quantization means for quantizing the moving image signal based on the quantization coefficient;
Coefficient control means for controlling the quantization coefficient for each macroblock,
The coefficient control means compares the activity of a macroblock to be quantized with the activity of surrounding macroblocks, and calculates a quantization coefficient for the macroblock to be quantized based on the comparison result. Encoding device. In a video encoding device that divides a screen in a matrix form in units of blocks (macroblocks) composed of a predetermined number of pixels and encodes a video signal in units of the divided macroblocks,
A quantization means for quantizing a moving image signal based on a quantization coefficient set in units of macroblocks;
Coefficient control that controls the quantization coefficient for each macroblock based on a predetermined arithmetic expression that reduces the quantization coefficient if the activity of the macroblock to be quantized is low and increases the quantization coefficient if the activity is high Means and
The coefficient control means is
The activity of the macroblock to be quantized is compared with the activities of the surrounding macroblocks, and it is determined whether the macroblock is at the boundary between the low activity area and the high activity area. If it is determined, the moving picture coding apparatus characterized by reducing the value of the quantization coefficient calculated based on the predetermined arithmetic expression. The coefficient control means is
The moving picture coding apparatus according to claim 2, wherein when a residual component obtained by motion estimation is equal to or less than a predetermined amount, the quantization coefficient is not reduced even for a macroblock at a boundary portion. The video image signal is encoded by dividing the inside of the screen into a matrix in units of blocks (macroblocks) composed of a predetermined number of pixels, and calculating and quantizing a quantization coefficient for each divided macroblock. In the video encoding method,
A moving picture coding method characterized by comparing the activity of a macroblock to be quantized with the activity of surrounding macroblocks and calculating a quantization coefficient for the macroblock to be quantized based on the comparison result. The video image signal is encoded by dividing the inside of the screen into a matrix in units of blocks (macroblocks) composed of a predetermined number of pixels, and calculating and quantizing a quantization coefficient for each divided macroblock. In the video encoding method,
Based on a predetermined arithmetic expression that reduces the quantization coefficient if the activity of the macroblock to be quantized is low and increases the quantization coefficient if the activity is high, the quantization coefficient is controlled for each macroblock,
The activity of the macroblock to be quantized is compared with the activities of the surrounding macroblocks, and it is determined whether the macroblock is at the boundary between the low activity area and the high activity area. If it is determined that the value is, the moving picture coding method characterized by reducing the value of the quantization coefficient calculated based on the predetermined arithmetic expression. 6. The moving picture coding method according to claim 5, wherein when a residual component obtained by motion estimation is equal to or less than a predetermined amount, the quantization coefficient is not reduced even for a macroblock at a boundary portion.

Images