JPH0750841A - Motion vector detector and its method - Google Patents

Abstract

PURPOSE:To detect an optimum motion vector by selecting the vector retrieval method in response to a monitor result of a monitor means so as to execute detection processing of the motion vector. CONSTITUTION:Picture element data and prediction data are inputted to a coding mode selection section 201 to compare the energy between the picture element value and a difference of the picture element and the predicted value and selects a coding mode according to coding mode control from a picture coding control section 214 and the compound result. The control section 214 discriminates a significant block, a quantization section 204 applies variable length coding to a quantized DCT coefficient and the result is outputted to a buffer section 206, in which the result is stored. The storage quantity is outputted to a motion vector detection section 211 and read from a frame memory section 210 by using the vicinity of the processing micro block position of a current frame as a retrieval area. The retrieval procedure is selected depending on the data storage quantity of the buffer section 206 to detect a motion vector. An optimum motion vector is detected by using the optimum procedure in response to the storage quantity.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting apparatus and method, and more particularly, to motion compensation interframe coding which is a compression coding method of moving image information, and more particularly to a motion vector detecting procedure. It is about.

[0002]

2. Description of the Related Art First, the principle of motion compensation interframe coding used for compression coding of moving image information will be briefly described.

When an object moves between frames, it is similar to the coding processing part (macroblock) not at the same position as the previous coded frame (reference frame) but at a position separated by the motion of the object. Structure will exist. Therefore, as shown in FIG. 4, the motion vector of the object is estimated from the previous frame and the current frame, and the difference with the similar block and the motion vector are encoded.

Since the prediction error value in motion-compensated interframe coding is smaller than the prediction error value in simple interframe coding, efficient coding is possible even for images with large motion.

For example, CCITT Recommendation H.264. In the H.261, for a macroblock of 16 pixels x 16 pixels,
A motion vector search area of 48 pixels x 48 pixels is used, and the horizontal and vertical components of the motion vector are +/- 1
It is defined as an integer within 5.

To detect a motion vector, the sum of absolute differences (or the sum of squared differences) between corresponding pixels is calculated for the processing block of the current frame and the prediction block of the reference frame, and the vector having the minimum value is defined as the motion vector. presume. As this procedure, a full search procedure for performing the sum of absolute difference calculation for all block patterns in the search area of the reference frame and a simple procedure such as a three-step search procedure shown in FIG. 5 are general.

The three-step procedure is a method of simply detecting a motion vector by a three-step procedure rather than performing a sum of absolute difference calculation for all blocks in the search area. That is, first, the same position block (X) and four blocks (A1 to
A4) is calculated, and the block with the minimum value (A
4) is obtained. Then, the block (A4) and four blocks (B1 to B4) four pixels away from the block are calculated to obtain the block (B1) having the minimum value.

Next, the block (B1) and the four blocks (C1 to C4) two pixels away from the block are calculated, and the vector for the block (C1) having the minimum value is set as the motion vector.

[0009]

Generally, the difference absolute value sum operation between block pixels requires a large amount of operation as compared with other image encoding processing. Therefore, in the full search procedure, although the detection is reliable, the total amount of calculation for calculating the motion vector becomes enormous. Sometimes exceeded.

As a result, the frame rate of the image may reach a ceiling, and smooth (moving smoothly) moving image communication may be hindered. Also, it is hard to say that full search is efficient for input images with small changes.

On the other hand, in the three-step search procedure,
Although the amount of calculation is drastically reduced, for example, if an incorrect block is selected in the first step, it may not be possible to return to the optimum search (erroneous motion vector detection may occur). There has been a problem that there is a danger with respect to an input image that changes rapidly.

In view of the above-mentioned problems, it is an object of the present invention to make it possible to detect an optimum motion vector without hindering the improvement of the frame rate of an image.

[0013]

A motion vector detecting device of the present invention is a motion vector detecting device for detecting a motion vector of a processing block with respect to a search area of a reference frame, and monitoring means for monitoring a code amount of image information. And a motion detecting means for performing a motion vector detecting process by switching the vector search method according to the monitoring result of the monitoring means.

The motion vector detecting method of the present invention is
A motion vector detecting method for detecting a motion vector of a processing block with respect to a search area of a reference frame, comprising a monitoring step of monitoring a code amount of image information, and a motion vector searching procedure according to a monitoring result obtained in the monitoring step. And a motion detecting step of performing a motion vector detecting process by switching.

[0015]

Since the present invention comprises the above technical means, the search method of motion vector detection is switched according to the code amount of the moving image information, so that efficient and effective moving image coding processing is possible. Become.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the motion vector detecting apparatus and method according to the present invention will be described below with reference to the drawings. Figure 1
It is a schematic block diagram which shows the structural example of the moving image coding apparatus which employ | adopted the motion vector detection system in a present Example.

In FIG. 1, reference numeral 201 denotes an encoding mode selection unit, which inputs pixel value data, prediction value data from a reference frame (previous frame), and a difference value (prediction error) between input pixel value data. The inter-frame difference value (prediction error value) with respect to the pixel value data input in macroblock units according to the energy comparison result between the two and the instruction of the image encoding control unit 214 (INTER mode), Input pixel values are encoded as they are (INTRA
Mode). Then, when the INTRA mode is selected, the input pixel value is output, and when the INTER mode is selected, the prediction error value is output.

Reference numeral 202 denotes a subtraction unit, which performs a subtraction process on the input pixel value data and the reference data. A DCT unit 203 performs a DCT (discrete cosine transform) process, which is a kind of orthogonal transform, on the input pixel value data. A quantization unit 204 selects a quantization step size according to an instruction from the image coding control unit 214, and inputs the DCT.
Quantize coefficient data.

Reference numeral 205 denotes a variable length coding unit, which performs variable length coding on the quantized DCT coefficient data. Two
Reference numeral 06 is a transmission buffer unit. Reference numeral 207 denotes an inverse quantization unit, which inversely quantizes the quantized DCT coefficient data. An IDCT unit 208 performs IDCT (inverse discrete cosine transform) processing of the input DCT coefficient data.

An adder 209 adds a predicted value to the data processed in the INTER mode. 21
Reference numeral 0 denotes a frame memory unit, which is a frame memory for motion-compensated interframe prediction. 211 is a motion vector detection unit according to the present invention. A motion compensation unit 212 selects pixel value data in the corresponding macroblock unit in the previous frame according to the motion vector detected by the motion vector detection unit 211.

A filter unit 213 is a low-pass filter for a block for which motion compensation interframe prediction is performed. Reference numeral 214 denotes an image coding control unit, which selects a coding mode, a quantization step size, a significant block determination, a frame based on a control signal from an external system control unit and a buffer storage amount of the transmission buffer unit 206. Various coding controls such as drop (frame skip) are performed.

Next, the operation of the image coding apparatus of the present embodiment constructed as above will be described. First, the pixel value data and prediction value data in macroblock units are input to the subtraction unit 202, and the difference value (prediction error) between the pixel value data and the prediction value data is output to the encoding mode selection unit 201.

In the coding mode selection unit 201, pixel value data and prediction value data are input, and energy comparison is performed between the input pixel value and the difference value (prediction error value) between the input pixel value and the prediction value. . Then, the coding mode is selected according to the energy comparison result and the coding mode control signal from the image coding control unit 214.

Then, the selected encoding mode is INTE
The input pixel value is output in the R mode, and the prediction error value is output to the DCT unit 203 in the INTRA mode. DCT section 2
In 03, the pixel value data from the coding mode selection unit 201 is subjected to DCT (discrete cosine transform) processing which is a kind of orthogonal transform in block units, and DCT coefficient data is output to the quantization unit 204.

In the quantizing unit 204, the image coding control unit 2
The input DCT coefficient is quantized by the step size selected according to the quantization characteristic control signal from 14. The variable length coding unit 205 performs significant block determination based on the coding control signal from the image coding control unit 214,
The quantized DCT coefficient is variable-length coded and output to the transmission buffer unit 206.

The transmission buffer unit 206 is composed of a buffer memory, buffers the variable length encoded data and outputs it to the demultiplexing / multiplexing unit in the subsequent stage, and stores the buffer storage amount in the image encoding control unit 214 and the motion vector detecting unit. Two
11 is sent.

The inverse quantizer 207 includes a quantizer 2
The DCT coefficient quantization output of No. 04 is input, and the quantization unit 20
Inverse quantization is performed using the quantization step size selected in 4, and DCT coefficients are output.

In the IDCT section 208, the inverse quantization section 207
The DCT coefficient output of 1 is subjected to IDCT (Inverse Discrete Cosine Transform) processing and output to the adder 209. In the addition unit 209, I
For the macroblock data processed in the NTER mode, the corresponding prediction data is added to the output of the IDCT unit 208 and output. Further, for the data of the macroblock processed in the INTRA mode, the IDCT unit 2
The output of 08 is output as it is.

The frame memory unit 210 is a frame memory for motion-compensated interframe prediction, and is composed of at least two frame memories. And
The pixel value output of the addition unit 209 is accumulated, and at the same time, the pixel value data of the previous frame is output according to an instruction from the motion compensation unit 212 for motion compensation interframe prediction.

In the motion vector detecting unit 211, the previous frame pixel value data near the processing macroblock position of the current frame is read from the frame memory unit 210 as a motion vector search window (search area), and the data storage amount of the transmission buffer unit 206 is read. Block matching calculation is performed by a search procedure according to the above to detect a motion vector, and the detected motion vector is output to the motion compensation unit 212.

The motion compensation unit 212 reads out pixel value data of the corresponding macroblock of the previous frame from the frame memory unit 210 according to the motion vector detected by the motion vector detection unit 211 and outputs it.

The filter unit 213 is a low-pass filter provided for the purpose of alleviating the discontinuity at the block boundary due to the motion compensation, and performs a filtering process on the motion-compensated data to predict. The value data is output to the subtraction unit 202.

The image coding control unit 214 controls the timing of the entire image coding unit, and at the same time, the quantization in the quantizing unit 204 is adaptively performed by the quantization characteristic control signal based on the data storage amount of the transmission buffer unit 206. Selection of an encoding step size, control of a variable threshold, encoding mode determination control in an encoding mode selection unit 201 by an encoding mode control signal, variable length encoding unit 205 by an encoding control signal
Encoding control such as significant block determination or frame dropping (frame skipping) is performed.

Further, by constantly monitoring the image quality signal and various mode signals from the overall image coding unit or the external system control unit, it is possible to change the input image due to the change of the input means and the image quality setting of the user. Correspond.

Next, the structure and operation of the motion vector detecting section 211 will be described in detail. The motion vector is detected by calculating the sum of absolute differences (or the sum of squared differences) between corresponding pixels between the processing block of the current frame and the prediction block of the reference frame, and estimating the vector of the block having the minimum value as the motion vector. To do.

As described above, in general, the difference absolute value sum operation between block pixels requires a large amount of operation as compared with other image encoding processing. Therefore, in the case of the full search procedure of searching all the block patterns in the search area, although the detection is reliable, the total calculation amount for calculating the motion vector becomes enormous, and the upper limit of the frame rate of the image is set. It may occur. Moreover, it was hard to say that full search is efficient for an input image with a small change (small movement).

Here, since the motion-compensated data encodes the difference value between the frames in the subsequent block, the generated code amount can be suppressed to be small for the input image with small change (small motion). It is common.

When the generated code amount is suppressed to a small value,
The permissible time for the coding operation per frame is reduced, and conversely, the increase in the amount of generated code means the increase in the data transmission time, so that the permissible time for the coding operation is increased.

Therefore, in this embodiment, the calculation time is taken into consideration for an image with a small amount of generated codes, and
Moreover, the motion vector detection processing is performed using a search procedure by a simple method that does not hinder the improvement of the frame rate. For an image with a large amount of generated code, motion vector detection processing is performed using a detailed search procedure such as full search within a region that can always be appropriately detected.

FIG. 2 is a schematic block diagram showing the structure of the motion vector detecting section. In FIG. 2, 101 is the same as in FIG.
It is a detection procedure selection unit that switches between the full search procedure and the simple search procedure according to the buffer storage amount input from the transmission buffer unit 206 shown in FIG.

Reference numeral 102 denotes an operation execution unit, which is the detection control unit 1
This is for calculating the sum of absolute differences between corresponding pixels for the processing macroblock and the prediction block input from 04. A minimum determination unit 103 is for accumulating the calculation result of the calculation execution unit 102 and outputting to the detection control unit 104 a vector corresponding to a block having the minimum calculation result according to an instruction from the detection control unit 104. is there.

Reference numeral 104 denotes a detection control section, which inputs pixel value data of a processing macroblock, pixel value data of a preceding frame search area near the processing macroblock position from the frame memory section 211, and a selection procedure of the detection procedure selection section 101. Based on the above, the calculation execution unit 102 is instructed to transmit the corresponding data and to execute the calculation, and the minimum determination unit 103 is instructed to perform the minimum determination based on the selection procedure, and the obtained motion vector is controlled to be transmitted.

Next, the operation of the motion vector detecting section shown in FIG. 2 will be described. The detection procedure selection unit 101 compares the buffer storage amount from the transmission buffer unit 206 with a preset reference value. If the buffer storage amount exceeds the reference value, a detailed search procedure such as a full search procedure is performed. select. If it is less than or equal to the reference value, a simple search procedure such as a three-step procedure is selected, and the selection result is notified to the detection control unit 104.

The detection control unit 104 sends the processing block and the prediction block to the operation execution unit 102 according to the selection procedure, and instructs the operation execution. Further, the minimum determination unit 103 is instructed to perform the minimum determination operation.

Specifically, in the full search procedure,
Prediction data is sent to the calculation execution unit 103 for all block patterns together with the processing block data, the minimum value determination unit 103 is instructed to make the minimum calculation value determination for all block patterns, and the detected motion vector is calculated based on the determination result. It is output to the motion compensation unit 212.

In the simple search procedure (here, three-step search), the operation of the operation executing unit 102 of the nine blocks and the minimum value determination by the minimum value determining unit 103 are repeatedly executed and detected as in the above-mentioned procedure of FIG. Output motion vector.

FIG. 3 is a flow chart for explaining the above operation. First, in step S1, it is determined whether the generated code amount is equal to or larger than a reference value. Then, when the generated code amount is equal to or larger than the reference value, the process proceeds to step S2, and the motion vector detection process by the full search procedure is performed. If the generated code amount is less than or equal to the reference value, the process proceeds to step S3, and motion vector detection processing is performed by the simple search procedure.

By the above operation, when the input image has little motion and the generated code amount is kept small, the motion vector is detected by the simple search procedure. this is,
This is because the full search procedure is not only efficient, but also requires an enormous amount of calculation time, which may hinder further improvement of the frame rate.

On the other hand, when the change in the input image is large and the generated code amount exceeds the reference value, the calculation amount (time) does not become an obstacle, so that the full search procedure that can always perform appropriate detection. Is used to detect the motion vector.

As described in detail above, by switching the search procedure of motion vector detection according to the generated code amount after the compression of the moving image information, it is possible to optimize according to the generated code amount without hindering the improvement of the frame rate. Various motion vectors can be detected, and efficient and effective moving image coding processing can be performed.

It should be noted that, in the above embodiment, the motion vector search procedure is divided into a full search procedure and a simple search procedure.
Although only the case of the street selection is described, it is obvious that the case of changing the size of the search area, switching the search steps, or selecting a plurality of search procedures is also applicable. Further, the method of the simple search procedure is not limited to the above embodiment.

Further, in the above embodiment, only the case where the storage amount of the buffer is used as the selection criterion of the procedure has been described, but the switching according to the generated code amount, frame rate, etc. for one frame or several frames is not applicable. Needless to say. The present invention can also be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0053]

As described above, the present invention monitors the code amount of image information when detecting a motion vector of a processing block for a search area of a reference frame, and a motion vector search method according to the monitoring result. switching,
Since the motion vector detection process is performed, for example, when the code amount of the data is large, the motion detection process is performed by a detailed search procedure such as full search to perform an appropriate motion vector detection. When the code amount of data is small, the amount of calculation can be reduced and the upper limit of the frame rate can be increased by performing the motion detection process by a simple search procedure. That is, the present invention enables effective motion vector detection adapted to the code amount.

[Brief description of drawings]

FIG. 1 is a schematic block diagram showing a configuration example of a moving picture coding apparatus showing an embodiment of the present invention.

FIG. 2 is a schematic block diagram showing the configuration of a motion vector detection unit.

FIG. 3 is a flowchart illustrating an operation of a motion vector detection unit.

FIG. 4 is a diagram showing how a motion vector of an object is estimated from a previous frame and a current frame.

FIG. 5 is an explanatory diagram of a 3-step search procedure.

[Explanation of symbols]

 101 Detection Procedure Selection Unit 102 Operation Execution Unit 103 Minimum Judgment Unit 104 Detection Control Unit

Claims (2)

[Claims] 1. A motion vector detection device for detecting a motion vector of a processing block with respect to a search area of a reference frame, the monitoring means monitoring a code amount of image information, and a vector search according to a monitoring result of the monitoring means. A motion vector detecting device, comprising: a motion detecting unit that switches a method to perform a motion vector detecting process. 2. A motion vector detecting method for detecting a motion vector of a processing block with respect to a search area of a reference frame, comprising a monitoring step of monitoring a code amount of image information, and a monitoring result obtained in the monitoring step. And a motion detecting step of performing a motion vector detecting process by switching a motion vector search procedure.