JP2007274366A - Video encoding apparatus and motion vector detection method - Google Patents

Abstract

A motion vector detection method for coding a moving image that has good coding efficiency and that does not deteriorate the image quality even when coding an image with intense motion or the like.
Based on processing target image data, a motion vector of a macroblock is detected, a motion vector of a block obtained by dividing the macroblock is sequentially detected, a cumulative sum of differences corresponding to the motion vector of the block is obtained, and a macro Depending on the difference corresponding to the block motion vector, the cumulative sum of the differences corresponding to the block motion vector, and a predetermined threshold, the motion vector of the macroblock or the motion vectors of a plurality of blocks detected in sequence are encoded into a moving image. Adopt as a motion vector to be used.
[Selection] Figure 2

Description

  The present invention relates to a moving picture coding apparatus, and more particularly to a motion vector detection method in an MPEG-4 encoder.

  MPEG-4 has been established as a standard for moving picture encoding, and a small and high-performance MPEG-4 encoder for encoding moving pictures in accordance with the standard has been commercialized. Today, as shown in FIG. 6, the number of devices capable of recording moving images is increasing by installing an MPEG-4 encoder. The recorded moving image data can be transmitted to the playback side with higher efficiency and higher quality than conventional standards via a network or a recording medium such as a memory card.

  Not only MPEG-4 but also moving picture coding uses a correlation between picture frames to reduce redundancy in the time direction in order to compress the amount of information. Therefore, as described in the following patent document, a motion vector is detected for each 16 × 16 pixel region (macroblock) or each 8 × 8 pixel region (block) obtained by dividing the region, The amount of information required for encoding is compressed by using the detected motion vector data.

The data of the current frame of the moving image is encoded as difference data between the motion vector data and the reference frame. If the transmission rate or the available bandwidth is constant, the total amount of data for motion vectors and the amount of data for difference data is constant. The amount of data that can be used for encoding differential data is reduced.
JP 2005-278156 A

  In motion vector detection at the time of moving picture encoding, if one motion vector is detected and used for each macroblock, the number of bits used as a vector is reduced and the encoding efficiency is often improved. However, in the case of encoding an image with intense motion or an image with a fine image pattern, it is difficult to express the fine motion, and the image quality deteriorates.

  Accordingly, an object of the present invention is to realize a motion vector detecting means that has good encoding efficiency and that does not deteriorate the image quality even when an image with intense motion is encoded.

  According to the motion vector detection method for moving picture coding of the present invention, a motion vector of a macroblock is detected, a difference corresponding to the motion vector of the macroblock is obtained, and the macroblock to be processed is divided into a plurality of parts. For each of the divided blocks, a motion vector having the smallest difference from the block of the reference frame is sequentially detected, a cumulative sum of differences corresponding to the sequentially detected motion vectors is calculated, and a difference corresponding to the motion vector of the macroblock is calculated. Based on the cumulative sum of the differences and a predetermined threshold value, a motion vector of a macroblock or a plurality of motion vectors for each divided block is employed as a motion vector used for moving picture coding.

  In addition, according to the present invention, the moving picture coding apparatus detects a macroblock motion vector and obtains a difference corresponding to the motion vector, and divides the macroblock to be processed into a plurality of parts. For each divided block, a block motion vector detection unit that sequentially detects a motion vector having the smallest difference from the block of the reference frame, and a cumulative sum of differences corresponding to the motion vectors sequentially detected by the block motion vector detection unit are calculated. A difference accumulation sum calculation unit, a motion vector detection control unit, and a threshold storage unit are provided, and the motion vector detection control unit includes a difference corresponding to a motion vector detected by the macroblock motion vector detection unit, and a difference accumulation sum calculation unit. Based on the cumulative sum of the differences to be calculated and the threshold set in the threshold storage unit, the macroblock motion vector Out portion is to adopt as a motion vector to be used for video encoding the plurality of motion vectors detected by the motion vector or block motion vector detecting section detects.

  Further, the present invention has a surface as a program that implements the above-described moving picture coding apparatus as a computer and causes the computer to execute the motion vector detection method for moving picture coding.

  According to the present invention, a motion vector for a macroblock of 16 × 16 pixels and a motion vector for a block obtained by dividing the macroblock can be adopted as a motion vector used for video encoding according to a video sequence. In a normal image, one motion vector is used for a macroblock to increase the coding efficiency. When coding an image with intense motion or an image with a fine picture pattern, By using a plurality of motion vectors, image quality degradation can be avoided.

  FIG. 1 is a functional block diagram of an MPEG-4 encoder according to the present invention. A moving image to be encoded is input to the input image memory 21, and an image encoded only with pixel data in one frame is input to the DCT conversion unit 31 as it is. In the case of encoding using the reference frame pixel data stored in the reference frame memory 22, a motion vector is detected by the motion vector detection unit 10, and the motion compensation unit 11 performs motion using the motion vector. The difference from the compensated image frame is calculated by the difference calculation unit 81 and input to the DCT conversion unit 31.

  In the motion vector detection method of the present invention, in addition to the 16 × 16 pixel unit difference square sum stored in the macroblock difference memory 23, the 8 × 8 pixel unit difference square stored in the block difference cumulative sum memory 24 is used. The sum of the sums (cumulative sum) is used.

  The part other than the motion vector detection is the same as the conventional one, the DCT conversion unit 31 converts the time domain data into the frequency domain data, the coefficient data in the frequency domain is quantized by the quantization unit 41, On the other hand, the bit stream which is input to the variable length encoding unit 60 including the one via the AC / DC prediction unit 50 and is the final output is stored in the bit stream memory 26, for example. On the other hand, the quantized data is inversely transformed through an inverse quantization unit 42 and an inverse DCT transform unit 32, added by motion compensation data and an adder 82, and stored in the frame memory 25. Used for motion compensation in 11.

The bit stream information is fed back to the rate control unit 70 and used for rate control of the quantization unit 41 and the inverse quantization unit 42.
FIG. 2 is a functional block diagram of a motion vector detection portion in the moving picture coding apparatus of the present invention, and explains the principle of the present invention. Compared with the functional block diagram of the MPEG-4 encoder shown in FIG. 1, the input image memory 21 and the reference frame memory 22 shown in FIG. 1 can be included in the processing target image data storage unit 100 shown in FIG. Of course, instead of using the input image memory 21 and the reference frame memory 22 directly, it is possible to use buffers each having a capacity of several frames. Also, the macroblock motion vector detection unit 200, the block motion vector detection unit 300, the difference accumulation sum calculation unit 400, the motion vector detection control unit 500, and the threshold storage unit 600 in FIG. 2 correspond to the motion vector detection unit in FIG. To do. The macroblock motion vector detection unit 200 includes the macroblock difference memory 23 of FIG. 2, and the difference accumulation sum calculation unit 400 includes the block difference accumulation sum memory 24 of FIG.

  The processing target image data storage unit 100 stores image data of the current frame to be encoded and its reference frame. Based on the image data stored in the processing target image data storage unit 100, the macroblock motion vector detection unit 200 detects the motion vector of the macroblock having the smallest difference from the macroblock of the reference frame. There are various macroblock motion vector detection methods, but the present invention is not limited to a particular one of the macroblock motion vector detection methods.

  The block motion vector detection unit 300 divides the macro block to be processed into a plurality of blocks, and sequentially detects a motion vector having the smallest difference from the block of the reference frame for each of the divided blocks. Although there are various types of motion vector detection methods for blocks obtained by dividing a macroblock, as in the case of macroblocks, the present invention is limited to a specific one of block motion vector detection methods. It is not something.

The difference accumulation sum calculation unit 400 calculates the accumulation sum of differences corresponding to the motion vectors sequentially detected by the block motion vector detection unit 300.
The motion vector detection control unit 500 sets the difference corresponding to the motion vector detected by the macroblock motion vector detection unit 200, the cumulative sum of the differences calculated by the difference cumulative sum calculation unit 400, and the threshold set in the threshold storage unit. Based on this, a motion vector detected by the macroblock motion vector detection unit 200 or a plurality of motion vectors detected by the block motion vector detection unit 300 is adopted as a motion vector used for video encoding.

  FIG. 3 shows a more specific configuration example of the functional block shown in FIG. The current frame storage unit 110, in-process macroblock storage unit 120, and reference frame storage unit 130 in FIG. 3 constitute the processing target data storage unit 100 in FIG. The first difference calculation processing unit 210 and the macroblock difference storage unit 220 in FIG. 3 constitute a macroblock motion vector detection unit 200. The second difference calculation processing unit 310 and the block difference storage unit 320 constitute a block motion vector detection unit 300, and the cumulative addition unit 410 and the block difference accumulation sum storage unit 420 constitute a difference accumulation sum calculation unit 400.

Hereinafter, embodiments of the present invention will be described with reference to FIGS.
The current frame storage unit 110 in FIG. 3 stores the current frame shown in FIG. 4, that is, the image frame currently being processed. Further, the reference frame storage unit 130 stores a reference frame shown in FIG. 4, that is, an image frame referred to for encoding the current frame. The currently processed macroblock storage unit 120 stores the currently processed macroblock among the macroblocks that are 16 × 16 pixel areas in the current frame.

  The first difference calculation processing unit 210 is a part that performs motion vector detection in units of macroblocks, and includes a macro block stored in the processing-in-process macroblock storage unit 120 and a reference frame stored in the reference frame storage unit 130. A difference sum of squares is sequentially calculated as a difference from the 16 × 16 pixel area within the search range, and a motion vector is obtained by obtaining an area having the smallest difference square sum. Then, the smallest sum of squared differences is set in the macroblock difference storage unit 220. The first difference calculation processing unit 210 stores the difference square sum obtained in the first calculation in the macroblock difference storage unit 220 and then calculates the difference between the calculation result and the macroblock difference storage unit 220 every time the difference square sum is calculated. And the smaller one is stored in the macroblock difference storage unit 220, so that the smallest difference square sum value can be set in the macroblock difference storage unit 220.

  The second difference calculation processing unit 310 divides the currently processed macroblock stored in the currently processed macroblock storage unit 120 into four blocks B1, B2, B3, and B4 as shown in FIG. The motion vectors MV1, MV2, MV3, and MV4 that minimize the sum of squared differences are detected by sequentially calculating the sum of squared differences from the 8 × 8 pixel area within the reference frame search range stored in the reference frame storage unit 130. Each time a motion vector is detected, the sum of squared differences with the block in the corresponding reference frame is stored in the block difference storage unit 320, and the value is input to the cumulative addition unit 410. The second difference calculation processing unit 310 obtains a motion vector using the block difference storage unit 320 in the same manner as the above-described method in which the first difference calculation processing unit 210 obtains a motion vector using the macroblock difference storage unit 220. Method can be adopted.

  The cumulative sum of the difference squares when the motion vectors MV1, MV2, MV3, and MV4 are obtained by the cumulative addition unit 410 and the block difference cumulative sum storage unit 420 are sequentially obtained in the block difference cumulative sum storage unit 420.

  The motion vector detection control unit 500 compares the difference square sum set in the macroblock difference storage unit 220 with the first threshold value X set in the threshold storage unit 600 by the threshold setting unit 700, and the difference square sum is calculated. If it is not larger than the first threshold value X, that is, if the difference square value is sufficiently small, the motion vector detected by the first difference calculation processing unit 210 is adopted for the encoding process.

  In the above comparison, if the sum of squared differences is larger than the first threshold value X, the operation of the second difference calculation processing unit 310 is started. Each time the second difference calculation processing unit 310 detects a motion vector for the block obtained by dividing the macroblock, the cumulative sum of the square sums of differences, that is, the value stored in the block difference cumulative sum storage unit 420 is calculated. Subtraction is performed from the value stored in the macroblock difference storage unit 220, and the difference is compared with the second threshold value Y set in the threshold value storage unit 600. If the difference is not larger than the second threshold Y as a result of the comparison, in other words, if the cumulative sum of the sum of squared differences of the blocks is not sufficiently small, the processing in the second difference calculation processing unit 310 is terminated at that stage. Then, the motion vector detected by the first difference calculation processing unit 210 is employed for the encoding process.

  If the difference is larger than the second threshold even if the difference is calculated by the cumulative sum of the sum of squares of all blocks, the motion vectors of all the blocks detected by the second difference calculation processing unit 310 are calculated. Adopted for encoding process.

The motion vector detection control unit 500 executes other processes necessary for motion vector detection, such as various initial settings, in addition to the processes described above.
If a plurality of motion vectors are used for one macroblock, it is possible to prevent deterioration of image quality in an image with intense motion, but the encoding efficiency is reduced compared to the case of one motion vector, Conversely, the image quality may deteriorate. Therefore, it is preferable that the first threshold value and the second threshold value are made variable as program parameters, for example, so that the image quality can be tuned so that a trade-off between the image quality and the coding efficiency is possible.

  For this purpose, a threshold setting unit 700 is provided in the embodiment shown in FIG. The threshold setting unit 700 can change the setting of the threshold set in the threshold storage unit. As specific means for setting the threshold, a method using the threshold as a program parameter and using an externally input threshold, or a method for dynamically setting the threshold according to the motion vector detection execution result in the motion vector detection control unit is used. Can be adopted.

Next, FIG. 6 shows a motion vector detection processing flow of the present invention.
Step S10 corresponds to the processing in the first difference calculation processing unit 210 described above. Further, Step S50 and Step S60 correspond to processing in the second difference calculation processing unit 310 and the cumulative addition unit 410. The other steps correspond to the control by the motion vector detection control unit 500.

  First, in step S10, a motion vector that minimizes the sum of squared differences in 16 × 16 pixel units is detected, and the squared difference sum in 16 × 16 pixel units is obtained. In step S20, it is determined whether the difference square sum of 16 × 16 pixels is larger than the threshold value X. If not, the process branches to step S90, and the motion vector detected in step S10 is adopted as a motion vector for encoding. If the determination in step S20 is yes, an initial value is set for the index assigned to the plurality of blocks obtained by dividing the macroblock in step S30, and the processes in steps S40 to S80 are sequentially repeated.

  In step S40, it is determined whether or not the processing for all blocks has been completed. If not completed, a motion vector that minimizes the sum of squared differences of the 8 × 8 pixel unit block is detected in step S50 to obtain the sum of squared differences, and in step S60, the difference corresponding to the motion vector obtained in step S50. Find the cumulative sum of squares.

  In step S70, the value obtained by subtracting the cumulative sum obtained in step S60 from the square sum of differences obtained in step S10 is compared with the threshold Y. If the threshold Y is smaller than the subtracted value, the index is updated and the process returns to step S40. . If the threshold Y is not smaller than the subtracted value, the process branches to step S90.

  If it is determined in step S40 that the processing for all blocks has been completed, the process branches to step S100, and the four motion vectors detected in step S50 are employed as motion vectors for encoding. .

  Note that each arithmetic processing unit, control unit, and the like shown in FIG. 3 can be configured as a part having individual hardware, or configured as a computer that is a collective system including one or more CPUs. You can also. It is also possible to configure on the hardware configuring the entire MPEG-4 encoder. In any case, for those skilled in the art, such as a combination of hardware and software, the configuration in which the present invention is specifically realized can be variously selected based on the matters disclosed in this specification. It is a matter.

  In the description of the above embodiment, the motion vector for the block obtained by dividing the macroblock is detected after detecting the motion vector of the macroblock. However, by increasing the hardware, both motion vectors are detected. It will be apparent to those skilled in the art that the detection can be parallelized.

  In addition, the motion vector for the block obtained by dividing the macroblock first is sequentially detected, and if the cumulative difference sum does not exceed a certain threshold, a plurality of motion vectors for the block are adopted, and the cumulative difference sum is constant. If the threshold value is exceeded, the motion vector detection control unit may control to stop the macroblock division and switch to macroblock motion vector detection.

  Furthermore, it is also possible to perform a process of switching which motion vector detection is executed first depending on the frequency at which the motion vector of the macroblock is adopted and the frequency at which the motion vector of the divided block is adopted. . In these cases as well, it is possible to modify the predetermined threshold value as a program parameter.

(Appendix 1)
A macroblock motion vector detector that detects a motion vector of the macroblock and obtains a difference corresponding to the motion vector;
A block motion vector detection unit that divides the macroblock to be processed into a plurality of blocks, and sequentially detects a motion vector having the smallest difference from the block of the reference frame for each divided block;
A cumulative difference calculation unit that calculates a cumulative sum of differences corresponding to motion vectors sequentially detected by the block motion vector detection unit;
A motion vector detection control unit;
A threshold storage unit;
The motion vector detection control unit includes a difference corresponding to a motion vector detected by the macroblock motion vector detection unit, a cumulative sum of differences calculated by the difference cumulative sum calculation unit, and a threshold set in the threshold storage unit Based on the above, a motion vector detected by the macroblock motion vector detection unit or a plurality of motion vectors detected by the block motion vector detection unit is adopted as a motion vector used for video encoding. Image encoding device.
(Appendix 2)
A current frame storage unit that stores a current frame that is a currently processed image frame; a reference frame storage unit that stores a reference frame that is an image frame referred to for encoding the current frame; A processing-in-progress macroblock storage unit that stores a macroblock being processed among the macroblocks,
The macroblock motion vector detection unit includes a first difference calculation processing unit and a macroblock difference storage unit,
The first difference calculation processing unit calculates a difference square as a difference between a macroblock stored in the processing-in-process macroblock storage unit and a region of 16 × 16 pixels within a reference frame search range stored in the reference frame storage unit. The motion vector is obtained by sequentially calculating the sum and obtaining the smallest area of the sum of squared differences, and the value of the smallest sum of squared differences is set in the macroblock difference storage unit,
The block motion vector detection unit includes a second difference calculation processing unit and a block difference storage unit, and the difference cumulative sum calculation unit includes a cumulative addition unit and a block difference cumulative sum storage unit,
The second difference calculation processing unit divides a macroblock stored in the processing-in-process macroblock storage unit into a plurality of blocks, and blocks within a reference frame search range stored in the reference frame storage unit for each block Each time the difference square sum is calculated and the smallest difference square sum is stored in the block difference storage unit, the motion vector of the block is detected, and each time the motion vector of the block is detected, the block difference The sum of squared differences with the block in the reference frame corresponding to the motion vector stored in the storage unit is input to the cumulative addition unit,
The cumulative addition unit stores the sum of the input difference square sum and the block difference cumulative sum stored in the block difference cumulative sum storage unit as a new block difference cumulative sum in the block difference cumulative sum storage unit,
The threshold storage unit stores a first threshold and a second threshold,
The motion vector detection control unit compares the difference square sum set in the macroblock difference storage unit with the first threshold value set in the threshold value storage unit, and the difference square sum is greater than the first threshold value. If not, the motion vector detected by the first difference calculation processing unit is adopted. If the difference sum of squares is larger than the first threshold in the comparison, the operation of the second difference calculation processing unit is performed. Each time the second difference calculation processing unit detects a motion vector for a block obtained by dividing a macroblock, the block difference cumulative sum stored in the block difference cumulative sum storage unit is stored in the macroblock difference storage unit. Subtract from the stored sum of squared differences, compare the difference of the subtraction results with a second threshold set in the threshold storage unit, and if the difference is not greater than the second threshold Even if the motion vector detected by the first difference calculation processing unit is adopted, and the difference of the subtraction result is calculated by the cumulative sum of the sum of squares of all the blocks, the difference is larger than the second threshold value. In such a case, the moving image encoding apparatus according to appendix 1, wherein the motion vectors of all the blocks detected by the second difference calculation processing unit are adopted as motion vectors used for moving image encoding.
(Appendix 3)
The moving picture encoding apparatus according to attachment 2, further comprising a threshold setting unit that sets the first threshold or the second threshold.
(Appendix 4)
In a motion vector detection method for video encoding,
Detecting a motion vector of the macroblock and obtaining a difference corresponding to the motion vector of the macroblock;
Dividing the macroblock to be processed into a plurality of
For each divided block, sequentially detect the motion vector with the smallest difference from the reference frame block,
Calculating a cumulative sum of differences corresponding to the sequentially detected motion vectors;
Based on the difference corresponding to the motion vector of the macroblock, the cumulative sum of the differences, and a predetermined threshold, the motion vector of the macroblock or a plurality of motion vectors for each of the divided blocks is encoded as a moving image. A motion vector detection method characterized in that it is adopted as a motion vector to be used in the operation.
(Appendix 5)
The motion vector detection method according to appendix 4, wherein the predetermined threshold can be changed.
(Appendix 6)
The predetermined threshold values are a first threshold value and a second threshold value,
If the difference corresponding to the motion vector of the macroblock is not greater than the first threshold, the motion vector of the macroblock is adopted as a motion vector to be used for video encoding;
When the difference corresponding to the motion vector of the macroblock is greater than the first threshold, the cumulative sum of the difference is calculated each time the difference sum corresponding to the motion vector of the sequentially detected block is calculated. Subtracting from the difference corresponding to the motion vector of the macroblock, comparing the difference of the subtraction result with the second threshold,
If the difference is not larger than the second threshold, the motion vector of the macroblock is adopted as a motion vector used for video encoding;
Even if the difference of the subtraction result is calculated by the cumulative sum of the differences for all blocks, if the difference is larger than the second threshold value, the motion vectors of all the blocks are used for moving picture coding. The motion vector detection method according to appendix 4, which is employed as a motion vector.
(Appendix 7)
The motion vector detection method according to appendix 6, wherein the first threshold value and the second threshold value can be changed.
(Appendix 8)
Divide the macroblock to be processed into multiple blocks,
For each divided block, sequentially detect the motion vector with the smallest difference from the reference frame block,
Calculating a cumulative sum of differences corresponding to the motion vectors of the sequentially detected blocks;
Comparing the cumulative sum of the sequentially calculated differences with a predetermined threshold;
If the cumulative sum of the differences is greater than the predetermined threshold, the motion vector of the macroblock is detected and used as a motion vector used for video encoding;
Even if the cumulative sum of differences for all blocks is calculated, if the cumulative sum of the differences is not greater than the predetermined threshold, the motion vectors of all the blocks are adopted as motion vectors used for video encoding. The motion vector detection method according to supplementary note 4, wherein:
(Appendix 9)
The motion vector detection method according to appendix 8, wherein the predetermined threshold can be changed.
(Appendix 10)
In a program for causing a computer to perform motion vector detection for video encoding,
On the computer,
A macroblock motion vector detection step for detecting a motion vector of the macroblock and obtaining a difference corresponding to the motion vector of the macroblock;
Dividing the macroblock to be processed into a plurality of parts;
A block motion vector detection step for sequentially detecting a motion vector having the smallest difference from the block of the reference frame for each divided block;
A cumulative sum calculation step of calculating a cumulative sum of differences corresponding to the motion vectors of the blocks sequentially detected in the block motion vector detection step;
The macroblock motion vector detection based on the difference corresponding to the motion vector of the macroblock obtained in the macroblock motion vector detection step, the cumulative sum of the differences calculated in the cumulative sum calculation step, and a predetermined threshold value A motion vector adopting step that employs a motion vector of the macroblock detected in the step or a plurality of motion vectors for each block detected in the block motion vector detecting step as a motion vector used for video encoding is executed. A program characterized by that.
(Appendix 11)
The program according to claim 10, wherein the predetermined threshold is a program parameter.
(Appendix 12)
The computer includes a current frame storage unit that stores a current frame that is a currently processed image frame, and a reference frame storage unit that stores a reference frame that is an image frame referred to for encoding the current frame. A processing-in-progress macroblock storage unit that stores a macroblock being processed among the macroblocks of the current frame, a macroblock difference storage unit, a block difference storage unit, a block difference cumulative sum storage unit, a first threshold value, A threshold storage unit storing the second threshold,
The macroblock motion vector detecting step includes calculating a square of a difference as a difference between a macroblock stored in the processing-in-process macroblock storage unit and a 16 × 16 pixel region within a reference frame search range stored in the reference frame storage unit. The motion vector of the macro block is detected by calculating the sum by sequentially calculating the motion vector by finding the area with the smallest sum of squared differences, and setting the value of the smallest sum of squared differences in the macro block difference storage unit. And obtaining a difference corresponding to the motion vector of the macroblock,
The block motion vector detection step calculates the sum of squared differences with respect to a block within a search range of a reference frame stored in the reference frame storage unit for each block, and calculates a minimum difference square sum to the block difference storage unit. To detect the motion vector of the block while storing
In the cumulative sum calculation step, each time a motion vector of the block is detected, the sum of squared differences with the block in the reference frame corresponding to the motion vector stored in the block difference storage unit and the block difference accumulation By storing the sum of the block difference cumulative sums stored in the sum storage unit as a new block difference cumulative sum in the block difference cumulative sum storage unit, the motion vector of the block sequentially detected in the block motion vector detection step is obtained. Which calculates the cumulative sum of the corresponding differences,
The motion vector adoption step compares the difference square sum set in the macroblock difference storage unit with the first threshold value set in the threshold value storage unit, and the difference square sum is larger than the first threshold value. If not, the motion vector detected by the first difference calculation processing unit is adopted as a motion vector used for moving image encoding. In the comparison, if the sum of squared differences is larger than the first threshold, The block difference accumulation stored in the block difference accumulation sum storage unit is started each time the second difference operation processing unit is started and the second difference calculation processing unit detects a motion vector for a block obtained by dividing the macroblock. Subtracting the sum from the sum of squared differences stored in the macroblock difference storage unit, comparing the difference of the subtraction results with a second threshold set in the threshold storage unit, If the difference is not larger than the second threshold value, the motion vector detected by the first difference calculation processing unit is adopted as a motion vector used for moving image encoding, and the sum of squared differences for all blocks is accumulated. Even if the difference of the subtraction result is calculated by summation, if the difference is larger than the second threshold value, the motion vectors of all blocks detected by the second difference calculation processing unit are used for moving image encoding. The program according to appendix 10, wherein the program is adopted as a motion vector to be executed.
(Appendix 13)
The program according to appendix 12, wherein the first threshold value and the second threshold value are program parameters.

2 is a functional block diagram of an MPEG-4 encoder according to the present invention. FIG. It is a functional block diagram of the motion vector detection part in the moving image encoder of this invention. 3 is a more specific configuration example of a functional block shown in FIG. 2. It is a figure which shows the motion vector detection of a macroblock unit. It is a figure which shows block division and vector detection of a block unit. It is a figure which shows the motion vector detection process flow of this invention. It is a figure which shows the example of the environment where an MPEG-4 encoder is used.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Motion vector detection part 11 Motion compensation part 21 Input image memory 22 Reference frame memory 23 Macroblock difference memory 24 Block difference accumulation sum memory 25 Frame memory 26 Bit stream memory 31 DCT conversion part 32 Inverse DCT conversion part 41 Quantization part 42 Inverse Quantization unit 50 AC / DC prediction unit 60 Variable length coding unit 70 Rate control unit 100 Processing target image data storage unit 110 Current frame storage unit 120 In-process macroblock storage unit 130 Reference frame storage unit 200 Macroblock motion vector detection unit 210 First difference calculation processing unit 220 Macroblock difference storage unit 300 Block motion vector detection unit 310 Second difference calculation processing unit 320 Block difference storage unit 400 Difference cumulative sum calculation unit 410 Cumulative addition unit 420 Block difference cumulative sum storage unit 00 motion vector detection control unit 600 threshold storage unit 700 threshold setting unit

Claims (5)

A macroblock motion vector detector that detects a motion vector of the macroblock and obtains a difference corresponding to the motion vector;
A block motion vector detection unit that divides the macroblock to be processed into a plurality of blocks, and sequentially detects a motion vector having the smallest difference from the block of the reference frame for each divided block;
A cumulative difference calculation unit that calculates a cumulative sum of differences corresponding to motion vectors sequentially detected by the block motion vector detection unit;
A motion vector detection control unit;
A threshold storage unit;
The motion vector detection control unit includes a difference corresponding to a motion vector detected by the macroblock motion vector detection unit, a cumulative sum of differences calculated by the difference cumulative sum calculation unit, and a threshold set in the threshold storage unit Based on the above, a motion vector detected by the macroblock motion vector detection unit or a plurality of motion vectors detected by the block motion vector detection unit is adopted as a motion vector used for video encoding. Image encoding device. A current frame storage unit that stores a current frame that is a currently processed image frame; a reference frame storage unit that stores a reference frame that is an image frame referred to for encoding the current frame; A processing-in-progress macroblock storage unit that stores a macroblock being processed among the macroblocks,
The macroblock motion vector detection unit includes a first difference calculation processing unit and a macroblock difference storage unit,
The first difference calculation processing unit calculates a difference square as a difference between a macroblock stored in the processing-in-process macroblock storage unit and a region of 16 × 16 pixels within a reference frame search range stored in the reference frame storage unit. The motion vector is obtained by sequentially calculating the sum and obtaining the smallest area of the sum of squared differences, and the value of the smallest sum of squared differences is set in the macroblock difference storage unit,
The block motion vector detection unit includes a second difference calculation processing unit and a block difference storage unit, and the difference cumulative sum calculation unit includes a cumulative addition unit and a block difference cumulative sum storage unit,
The second difference calculation processing unit divides a macroblock stored in the processing-in-process macroblock storage unit into a plurality of blocks, and blocks within a reference frame search range stored in the reference frame storage unit for each block Each time the difference square sum is calculated and the smallest difference square sum is stored in the block difference storage unit, the motion vector of the block is detected, and each time the motion vector of the block is detected, the block difference The sum of squared differences with the block in the reference frame corresponding to the motion vector stored in the storage unit is input to the cumulative addition unit,
The cumulative addition unit stores the sum of the input difference square sum and the block difference cumulative sum stored in the block difference cumulative sum storage unit as a new block difference cumulative sum in the block difference cumulative sum storage unit,
The threshold storage unit stores a first threshold and a second threshold,
The motion vector detection control unit compares the difference square sum set in the macroblock difference storage unit with the first threshold value set in the threshold value storage unit, and the difference square sum is greater than the first threshold value. If not, the motion vector detected by the first difference calculation processing unit is adopted. If the difference sum of squares is larger than the first threshold in the comparison, the operation of the second difference calculation processing unit is performed. Each time the second difference calculation processing unit detects a motion vector for a block obtained by dividing a macroblock, the block difference cumulative sum stored in the block difference cumulative sum storage unit is stored in the macroblock difference storage unit. Subtract from the stored sum of squared differences, compare the difference of the subtraction results with a second threshold set in the threshold storage unit, and if the difference is not greater than the second threshold Even if the motion vector detected by the first difference calculation processing unit is adopted, and the difference of the subtraction result is calculated by the cumulative sum of the sum of squares of all the blocks, the difference is larger than the second threshold value. In this case, the motion vector encoding apparatus according to claim 1, wherein motion vectors of all blocks detected by the second difference calculation processing unit are adopted as motion vectors used for video encoding. In a motion vector detection method for video encoding,
Detecting a motion vector of the macroblock and obtaining a difference corresponding to the motion vector of the macroblock;
Dividing the macroblock to be processed into a plurality of
For each divided block, sequentially detect the motion vector with the smallest difference from the reference frame block,
Calculating a cumulative sum of differences corresponding to the sequentially detected motion vectors;
Based on the difference corresponding to the motion vector of the macroblock, the cumulative sum of the differences, and a predetermined threshold, the motion vector of the macroblock or a plurality of motion vectors for each of the divided blocks is encoded as a moving image. A motion vector detection method characterized in that it is adopted as a motion vector to be used in the operation. The predetermined threshold values are a first threshold value and a second threshold value,
If the difference corresponding to the motion vector of the macroblock is not greater than the first threshold, the motion vector of the macroblock is adopted as a motion vector to be used for video encoding;
When the difference corresponding to the motion vector of the macroblock is greater than the first threshold, the cumulative sum of the difference is calculated each time the difference sum corresponding to the motion vector of the sequentially detected block is calculated. Subtracting from the difference corresponding to the motion vector of the macroblock, comparing the difference of the subtraction result with the second threshold,
If the difference is not larger than the second threshold, the motion vector of the macroblock is adopted as a motion vector used for video encoding;
Even if the difference of the subtraction result is calculated by the cumulative sum of the differences for all blocks, if the difference is larger than the second threshold value, the motion vectors of all the blocks are used for moving picture coding. The motion vector detection method according to claim 3, wherein the motion vector detection method is employed as a motion vector. Divide the macroblock to be processed into multiple blocks,
For each divided block, sequentially detect the motion vector with the smallest difference from the reference frame block,
Calculating a cumulative sum of differences corresponding to the motion vectors of the sequentially detected blocks;
Comparing the cumulative sum of the sequentially calculated differences with a predetermined threshold;
If the cumulative sum of the differences is greater than the predetermined threshold, the motion vector of the macroblock is detected and used as a motion vector used for video encoding;
Even if the cumulative sum of differences for all blocks is calculated, if the cumulative sum of the differences is not greater than the predetermined threshold, the motion vectors of all the blocks are adopted as motion vectors used for video encoding. The motion vector detection method according to claim 3.

Images