JPH10164585A - MOTION VECTOR DETECTION DEVICE AND IMAGE ENCODING DEVICE HAVING THE MOTION VECTOR DETECTION DEVICE - Google Patents

Abstract

(57) [Summary] [Problem] To detect an appropriate motion vector without hindering improvement of a frame rate of an image. SOLUTION: A detection procedure selection unit 201 selects a full search procedure if the determination result from an appropriate vector determination unit 203 exceeds a preset reference value, and selects a simple search procedure if the determination result is equal to or less than the reference value. The selected procedure is output to the detection control unit 204, and the detection control unit 204
2 is instructed to execute the operation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is used for motion-compensated inter-frame encoding when encoding image data, and detects a motion vector of a processing block corresponding to a search area of a reference frame. It concerns the device.

[0002]

2. Description of the Related Art First, the principle of motion-compensated interframe coding used for compression coding of moving picture information will be briefly described. If the subject in the image moves between frames, it is similar to the coding processing part (macroblock) not at the same position of the previous coded frame (reference frame) but at a position separated by the movement of the subject. Structure exists. Therefore, the motion vector of the subject is estimated from the previous frame and the current frame, and the difference between the similar block and the motion vector is encoded.

[0003] In detecting a motion vector, all blocks constituting a frame are searched for a processing block of a current frame and a prediction block of a reference frame (full search), and a sum of absolute differences (or a sum of squares of differences) between corresponding pixels is searched. ) Is calculated, and the vector having the minimum value is estimated as the motion vector.

[0004]

However, the sum of absolute differences between block pixels requires a large amount of calculation compared to other image coding processes. For this reason, the full search procedure is reliable in detection, but the total amount of calculation for calculating the motion vector is enormous, so when real-time processing is required for communication or the like, the speed required for calculation processing is transmitted. In some cases, we could not keep up with the processing speed. As a result, the frame rate of the image may reach a plateau, which may hinder smooth (smooth) moving image communication. Also, it is difficult to say that the full search procedure is efficient for an input image having a small change.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem, and has as its object to enable detection of an appropriate motion vector without hindering improvement in the frame rate of an image. I do.

[0006]

Means for Solving the Problems and Effects of the Invention The invention according to claim 1 for achieving this object is used for motion-compensated inter-frame coding, and is a processing block corresponding to a search area of a reference frame. A motion vector estimating device for estimating a destination block of a motion vector in a current reference frame to be processed by referring to at least a motion vector in the immediately preceding reference frame; Detection control means capable of detecting a motion vector in a movement destination block predicted by the means with priority, and performing detection in accordance with a simple search procedure using the detected motion vector as a motion vector in a current reference frame to be processed. A motion vector detection device characterized by comprising:

According to the motion vector detecting apparatus, when detecting the motion vector of the processing block corresponding to the search area of the reference frame, first, at least the motion vector of the immediately preceding reference frame is referred to to determine the current processing target. A destination block of a motion vector in a reference frame is predicted. The detection control means can execute the detection according to the simple search procedure. The detection according to the simple search procedure is to preferentially detect the motion vector in the predicted destination block, and the detected motion vector is subjected to the current processing. This is a motion vector in the target reference frame.

Since the amount of calculation required for detecting a motion vector is considerably large, if a motion vector is individually detected for all blocks in a reference frame, the total amount of calculation becomes enormous. Therefore, if the motion vector in the predicted destination block is the motion vector in the current reference frame to be processed, the motion vectors of the other blocks in the reference frame do not need to be individually calculated, and the motion vector The time required for the calculation can be reduced.

Conventionally, since the total amount of calculation for calculating a motion vector is enormous, when real-time processing such as communication is required, the speed required for calculation processing may not be able to keep up with the transmission processing speed, and the frame rate of an image may be reduced. In some cases, the smoothing of moving image communication may be hindered due to a plateau, but according to the present invention, the operation time of the motion vector is shortened, and such an inconvenience can be prevented. That is, an appropriate motion vector can be detected without hindering the improvement of the frame rate of the image.

Here, conventionally, the motion vectors are detected for all the blocks in the reference frame.
In the present invention, the reason why only the detection of a motion vector for one block is sufficient will be described. In the present invention, for example, as shown in FIG. 3 (A), when a certain block in a frame is moving at a constant speed, if only the motion vector Vn-1 of the immediately preceding reference frame is known, it is based on that. The destination block can be obtained quite accurately from the current motion vector Vn predicted in this way. That is, conventionally, FIG.
When searching for block C from block B in (A), after detecting motion vectors for all blocks in the area S1 indicated by the broken line, an optimum motion vector is first determined from the motion vectors for the reference frame. Was calculated. On the other hand, in the case of the present invention, the motion vector in the immediately preceding reference frame, that is, FIG.
Based on the motion vector Vn-1 from block A to block B in (A), for example, if moving at a constant speed, the directions and magnitudes of the vectors become equal, so that the motion vector Vn in the current reference frame is moved. The destination block can be predicted. As described above, when the movement has regularity, there is a high possibility that the motion vector detected in the movement destination block predicted in this way is the optimum motion vector in the reference frame.

As described above, in the present invention, a method of detecting according to the simple search procedure is employed in order to reduce the operation time of the motion vector. However, in all cases, the detection according to the simple search procedure is not preferable. In other words, this simple search procedure is a method of performing a search with priority given to the destination block obtained based on the “prediction”. Therefore, if a movement contrary to the prediction occurs, The motion vector in the block may not be appropriate. In the above-described example, an ideal state in which a certain block in a frame is moving at a constant speed is assumed. However, since not only such a movement is always performed, it is necessary to consider them.

In consideration of this point, a motion vector detecting device according to a second aspect is provided. That is, the configuration is such that the detection control means according to claim 1 can also execute detection according to a full search procedure for individually detecting a motion vector for all search regions in a reference frame.
Calculating means for calculating a predicted block of a reference frame based on a motion vector detected by executing detection according to a simple search procedure; a predicted block of a reference frame calculated by the calculating means; and a processing block of a frame currently being processed Calculating a sum of absolute differences or a sum of squares of differences between corresponding pixels in (a) and (b), determining whether the calculated value is appropriate; and determining that the calculated value is not appropriate by the determining means. And a detection procedure instructing means for instructing the detection control means to perform detection according to the full search procedure.

In this case, by performing the detection according to the simple search procedure, the prediction block of the reference frame is calculated based on the predicted motion vector of the movement destination block, and the prediction block of the calculated reference frame is calculated. A sum of absolute differences or a sum of squares of differences between corresponding pixels in the processing block of the frame currently being processed is calculated, and it is determined whether the calculated value is appropriate. If the calculated value is not appropriate, the control unit instructs the detection control unit to perform detection according to the full search procedure.

That is, based on the detection according to the simple search procedure, it is determined whether or not the detection result is appropriate, and if not, the detection is switched to the detection according to the full search procedure. Regardless of whether or not the range of motion vectors in the frame is appropriate, an appropriate motion vector in the reference frame can be detected as a result, and the time required for calculating the motion vector can be reduced as much as possible. .

[0015] In this case, the determination means determines the sum of absolute differences or the sum of squares of differences between corresponding pixels in the prediction block of the reference frame and the processing block of the frame currently being processed. It is conceivable to configure so as to judge that it is not appropriate when the value is equal to or more than a predetermined reference value.

As described above, the prediction by the prediction means, which is a feature of the present invention, refers to the motion vector of at least the immediately preceding reference frame, and refers to the destination block of the motion vector in the current reference frame to be processed. Is to predict. Therefore,
Only the motion vector in the immediately preceding reference frame may be referred to, or the motion vector in a reference frame earlier than that may be referred to.

However, it is more likely that more accurate prediction can be made by referring to motion vectors in a plurality of reference frames. For example, as shown in FIG. 3B, motion vectors (Vn-1 to Vn) in a plurality of continuous reference frames are provided.
If you look at -5), you may find that it moves like a rotary motion. In this case, in the case of a rotational motion, the curvature R or the like is known, and the destination block can be predicted fairly accurately according to the motion vector Vn obtained based on the curvature.

As shown in FIG. 3C, the motion vectors (Vn-1 to Vn-6) in a plurality of continuous reference frames are set.
In some cases, it can be seen that the movement direction is reversed up and down at a predetermined cycle. In this case, if the reversal period and the moving direction are known, it is possible to predict the destination block quite accurately. For example, the predicted motion vector Vn in the case shown in FIG.
Although it cannot be determined only from the immediately preceding motion vector Vn-1, in this case, referring to the past six motion vectors,
Because of the regularity, it can be expected that the motion vector will be the same as the motion vector Vn-4 four times before.

In other words, no matter how regular the movement is, if only the motion vector in the immediately preceding reference frame is referenced, the movement of the movement block is slightly shifted or the movement direction is greatly changed. In such a case, it is conceivable that the motion may shift considerably, so that the prediction of the motion cannot be determined. Therefore, it can be said that it is preferable to adopt a method of referring to motion vectors in a plurality of continuous reference frames from the viewpoint of being able to cope with various types of motion.

As an image coding apparatus provided with such a motion vector detecting device, a configuration as described in claim 4 can be considered. That is, the input video signal is
A data conversion unit that converts the pixel data into a rectangular area and outputs the pixel data; an orthogonal transformation unit that performs an orthogonal transformation on the pixel data of the rectangular area output from the data conversion unit; and an orthogonal transformation data output from the orthogonal transformation unit Is divided by a predetermined quantization step value and output as transformed coded data; andvariable-length coding means that converts the transformed coded data output from the quantizing means into a variable-length code and outputs the result. A dequantizer for multiplying the output from the quantizer by a quantization step value to generate orthogonal transform data, and performing an inverse orthogonal transform on the orthogonal transform data output from the inverse quantizer to obtain a rectangle. Inverse orthogonal transform means for outputting pixel data of the area, and differential data between the pixel data output from the inverse orthogonal transform means and the pixel data of the rectangular area output from the data transform means An image coding apparatus comprising: a motion prediction unit that outputs to an orthogonal transformation unit; wherein the motion prediction unit includes the motion vector detection device according to any one of claims 1 to 3. This is a characteristic configuration.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of an image coding apparatus conforming to the MPEG standard provided with a motion vector detecting apparatus according to an embodiment of the present invention. The present image encoding apparatus is configured to receive an analog video signal from the outside, encode the digital video signal accordingly, and encode and output the data in a state where the data amount is compressed. Video is realized by temporally arranging a plurality of still image data. In the case of this embodiment, for example, 1/30
One piece of still image data is displayed every second. In the present embodiment, one still image data is, for example,
It is represented by 04 pixels and 480 pixels vertically. In the present embodiment, this is divided into 16 × 16 pixels, which are defined as macroblocks. Accordingly, one still image is a set of 30 vertical macroblocks and 44 horizontal macroblocks.

As shown in FIG. 1, the present image encoding apparatus 1
Is a “data conversion unit” for receiving an input video signal, converting the same into digital data in pixel units, storing the converted digital data, and outputting pixel data in macroblock units, which are areas of 16 × 16 pixels. Image memory 101 and subtraction unit 1 that performs a subtraction process between input pixel value data and reference data
02 and a DCT unit 10 as "orthogonal transform means" for performing DCT (discrete cosine transform) processing which is a kind of orthogonal transform.
3 and a quantizing unit 10 as "quantizing means" for quantizing the DCT coefficient data input from the DCT unit 103.
4 and the DCT quantized by the quantization unit 104
A variable-length encoding unit 105 as a “variable-length encoding unit” that performs variable-length encoding on coefficient data, and a transmission buffer unit that stores data encoded by the variable-length encoding unit 105 106, an inverse quantization unit 107 as “inverse quantization means” for inversely quantizing the DCT coefficient data quantized in the quantization unit 104, and the inverse quantization unit 1
ID for the DCT coefficient data inversely quantized at 07
IDCT section 108 as “inverse orthogonal transform means” for performing CT (inverse discrete cosine transform) processing, and IDCT section 10
8, an addition unit 109 for adding motion compensation to the data returned to the image data before the IDCT processing, and a frame memory unit for storing an output from the addition unit 109 for motion inter-frame prediction. 110 and a motion vector detecting unit 1 corresponding to the “motion vector detecting device” of the present invention.
11, a motion compensator 112 for selecting the corresponding pixel value data of the macroblock in the previous frame according to the motion vector detected by the motion vector detector 111, and an image encoding controller 114. I have.

In this embodiment, the subtracting section 10
2, an adding unit 109, a frame memory unit 110,
The motion vector detecting section 111 and the motion compensating section 112 constitute "motion predicting means". Further, the image encoding control unit 114 selects an encoding mode, a quantization step size, and a significance based on a control signal from an external system control unit (not shown) and the buffer accumulation amount in the transmission buffer unit 106 described above. Block judgment,
Various coding controls such as frame dropping are performed. Therefore, the above-described quantization unit 104 selects a quantization step size according to the instruction from the image encoding control unit 114, and
The CT coefficient data is quantized, and the variable-length coding unit 105 performs significant block determination, frame dropping, and the like in accordance with an instruction from the image coding control unit 114.

Next, the operation of the image coding apparatus 1 according to the present embodiment having the above configuration will be described. Since the motion vector detecting section 111 has a feature in the motion vector detecting operation, a detailed description thereof will be given later, and the outline of the operation of the image encoding device 1 will be described here.

A video signal to be compressed is stored in the image memory 101 in a state converted into digital data in pixel units. In the present embodiment, since the compression unit is processed with a macroblock of 16 × 16 pixels as one unit, the data is rearranged in macroblock units and output. The pixel data in units of macro blocks, which is the output of the image memory 101, is subtracted from the subtractor 102 and the motion vector detector 111.
Sent to The prediction data from the motion compensation unit 112 is also input to the subtraction unit 102, and the difference data (prediction error) between the pixel data and the prediction data is output to the DCT unit 103.

The DCT unit 103 performs a DCT (Discrete Cosine Transform) process on the difference data output from the subtraction unit 102. At this time, the calculation is performed in predetermined block units (for example, every 8 × 8 pixels). In this embodiment, 1
A macroblock is obtained by dividing image data of a frame into 16 × 16 pixels. In this case, DCT is performed four times in one macroblock. In the DCT operation, an 8 × 8 coefficient is obtained as an output. In the present embodiment, this coefficient (hereinafter, also referred to as “DCT coefficient”).
Are output to the quantization unit 104 in the order of the frequency components.

In the quantization section 104, the image coding control section 1
The DCT coefficient input from the DCT unit 103 is quantized with the step size selected in accordance with the quantization characteristic control signal from the CPU 14, and the quantized value is input to the variable-length encoding unit 105.
Output to The variable-length coding unit 105 performs a significant block determination and the like based on the coding control signal from the image coding control unit 114, performs variable-length coding on the quantized DCT coefficient, and outputs it to the transmission buffer unit 106. In this variable-length coding, for example, it is conceivable that after the output of the quantization unit 104 is run-length-transformed, the coefficient is converted into a variable-length code and output as compressed data. The run-length encoding results in two coefficients, which are the set of the number of zero coefficients and the value of the next non-zero coefficient. However, the output of the quantization unit 104 tends to have many zero coefficients.
The data amount is reduced by run-length encoding. Next, two coefficients that are the result of the run-length encoding are put together, and one variable-length code is assigned and output, so that compressed data is obtained.

The transmission buffer unit 106 is composed of a buffer memory, buffers the variable-length coded data and outputs the buffered data to a separation / multiplexing unit (not shown) at the subsequent stage. Send out. on the other hand,
Since the inverse quantization unit 107 receives the quantization of the DCT coefficient output from the quantization unit 104, the inverse quantization is performed using the quantization step size selected by the quantization unit 104, The DCT coefficient is calculated and output to IDCT section 108. Since the result at this time is a value including an error due to quantization, IDCT section 108 applies IDCT to the DCT coefficient input from inverse quantization section 107.
(Inverse discrete cosine transform), and outputs the result to addition section 109.

The addition section 109 adds the output from the IDCT section 108 and the output from the motion compensation section 112 and outputs the result to the frame memory section 110. Frame memory unit 11
Reference numeral 0 denotes a frame memory for motion compensated inter-frame prediction, which is composed of at least two frame memories. Then, at the same time as accumulating the pixel value output from the adding unit 109, the pixel data of the previous frame is output to the motion vector detecting unit 111 and the motion compensating unit 112 for motion compensated frame prediction.

The motion vector detecting section 111 reads out the pixel data of the previous frame near the processing macroblock position of the current frame from the frame memory section 110 as a motion vector search window (search area).
A motion vector is detected by performing a block matching operation, and the detected motion vector is output to the motion compensation unit 112. Further, the motion vector detected by the block matching calculation is also output as a motion vector for an adjacent macro block.

In accordance with the motion vector detected by the motion vector detecting section 111, the motion compensating section 112 reads out the pixel data of the corresponding macroblock of the previous frame from the frame memory section 110 and outputs it to the subtracting section 102 as predicted value data. I do. This is the general operation of the image encoding apparatus 1, and as described above, efficient compression of data can be realized by employing, for example, run-length encoding and variable-length codes.

Next, the configuration and operation of the motion vector detecting section 111 will be described in detail with reference to FIG. First, a basic operation relating to motion vector detection and a characteristic portion of the present invention will be described. The motion vector is detected by calculating the sum of absolute differences or the sum of squares of the differences between the corresponding pixel data for 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. . As described above, generally, the sum of absolute differences between block pixel data requires a large amount of calculation as compared with other image encoding processes. For this reason, in the case of the “full search procedure” in which all block patterns in the search area are searched, the detection is reliable, but the total operation amount for calculating the motion vector becomes enormous, and the frame rate of the image is reduced. The upper limit of the allowable range may be exceeded. Also, the change is small (there is little movement)
Full search was not efficient for input images.

Therefore, in the motion vector detecting section 111 of the present embodiment, a simple method that is efficient and does not hinder the improvement of the frame rate in consideration of the calculation time for an image having a small generated code amount. (Hereinafter referred to as "simple search procedure") to perform the motion vector detection processing. And
When the search by the simple search procedure is not appropriate, the motion vector detection processing is always performed by using the search by the full search procedure that enables appropriate detection.

FIG. 2 is a block diagram schematically showing an internal configuration of the motion vector detecting section 111. As shown in FIG. 2, the motion vector detection unit 111 performs a full search procedure and a simple search procedure for the detection control unit 204 corresponding to the “prediction unit” and the “detection control unit”. A detection procedure selection unit 201 serving as a “detection procedure instruction unit” for instructing switching, and a “calculation unit” for calculating the sum of absolute differences between corresponding pixels for a processing macroblock and a prediction block input from the detection control unit 204 And an appropriate vector determination unit 203 as a “judgment unit” that outputs a vector corresponding to a block having the minimum calculation result to the detection control unit 204 according to an instruction from the detection control unit 204. ing.

The detection control unit 204 receives the pixel data of the processing macroblock from the frame memory unit 110 and the search area pixel data of the previous frame near the processing macroblock position from the image memory 101. On the other hand, the detection procedure selection unit 201 selects a full search procedure if the determination result from the appropriate vector determination unit 203 exceeds a preset reference value, and selects a simple search procedure if the determination result is less than the reference value. The detected procedure is output to the detection control unit 204.

The detection control unit 204 sends the data to be calculated to the calculation execution unit 202 and instructs execution of the calculation based on the procedure (full search procedure or simple search procedure) specified by the detection procedure selection unit 201. I do.
Further, it instructs the appropriate vector determination unit 203 to execute an appropriate vector determination based on the instructed selection procedure, and sends the motion vector obtained from the appropriate vector determination unit 203 to the motion compensation unit 112. Perform control. Note that the appropriate vector determination unit 203 is configured to transmit the determination result to the detection procedure selection unit 201 only when the motion vector is detected by the simple search procedure.

When detecting the motion vector of the processing block corresponding to the search area of the reference frame as detection according to the simple search procedure, first, at least the motion vector of the immediately preceding reference frame is referred to Predict the destination block of the motion vector in the frame. This prediction of the destination block is performed by the detection control unit 2
04 executes. Therefore, the detection control unit 204 stores at least the motion vector in the immediately preceding reference frame.

Then, the motion vector in the predicted destination block is detected with priority. In other words, the motion vector in the predicted block is detected even if it is not the first block when viewed from the predetermined search order in the reference frame to be processed. In particular,
When the detection control unit 204 sends the calculation target data to the calculation execution unit 202, the detection control unit 204 sends the calculation target data of the predicted movement destination block and instructs the execution of the calculation based on the simple search procedure.

Since this is a detection process according to the simple search procedure, the determination result is sent from the appropriate vector determination unit 203 to the detection procedure selection unit 201 as described above. If the determination result is equal to or greater than the reference value, the process proceeds to detection according to the full search procedure, and motion vectors for all blocks in the reference frame are detected. On the other hand, if the determination result is equal to or smaller than the reference value, it is determined that the detection result is appropriate, and is output to the motion compensation unit 112 as a motion vector.

Since the amount of calculation required for detecting a motion vector is considerably large, if a motion vector is individually detected for all blocks in a reference frame, the total amount of calculation becomes enormous. Therefore, if the motion vector in the predicted movement destination block can be used as the motion vector in the current reference frame to be processed, the motion vector does not need to be separately calculated for the other blocks in the reference frame, and the motion vector calculation can be performed. Can be shortened.

Conventionally, since the total amount of calculation for calculating a motion vector is enormous, when real-time processing such as communication is required, the speed required for calculation processing may not be able to keep up with the transmission processing speed, and the frame rate of an image may be reduced. In some cases, the smoothing of the moving image communication may be hindered due to a plateau. However, according to the detection according to the simple search procedure of the present invention, the operation time of the motion vector is reduced, and such inconvenience can be prevented.
That is, an appropriate motion vector can be detected without hindering the improvement of the frame rate of the image.

Here, the reason why the detection according to the full search procedure in which the motion vectors are detected for all the blocks in the reference frame in the related art is compared with the case where only the detection of the motion vector for one block is sufficient in the present invention. explain. That is, conventionally, when detecting a motion vector in a reference frame, the search order for a plurality of blocks constituting the reference frame has been uniformly determined according to the specifications of the apparatus. Therefore, after detecting the motion vectors for all the blocks, it is necessary for the first time to select the optimal motion vector from the motion vectors as the motion vector for the reference frame. The destination block of the motion vector in the current reference frame
This is because if the input image moves regularly, the motion vector detected in the movement destination block is likely to be the optimum motion vector in the reference frame.

That is, conventionally, when searching for the block C from the block B in FIG. 3A, after the motion vectors of all the blocks in the area S1 indicated by the broken line are detected, the optimum A motion vector with respect to a reference frame is calculated using a simple motion vector. On the other hand, in the present invention, based on the motion vector in the immediately preceding reference frame, that is, the motion vector Vn-1 from block A to block B in FIG. Since the directions and magnitudes are equal, the destination block of the motion vector Vn in the current reference frame can be predicted. in this way,
If the movement is regular, it is more likely that the motion vector detected in the movement destination block predicted in this way is the optimum motion vector in the reference frame, so that it can be handled by the simple search procedure. It will be.

In some cases, the detection according to the simple search procedure is not preferable. In such a case, in this embodiment, the simple search is performed based on the detection according to the simple search procedure which is effective in reducing the amount of calculation. The appropriate vector determination unit 203 determines whether the detection result obtained by the procedure is appropriate. If not, the detection is switched to the detection according to the full search procedure according to an instruction from the detection procedure selection unit 201. Regardless, the appropriate motion vector can be detected, and the time required for calculating the motion vector can be reduced as much as possible. In the case of detection according to the full search procedure in this case, detection is performed in the order of blocks according to a predetermined search order, as in the related art.

As described above, the present invention is not limited to such an embodiment, and can be implemented in various forms without departing from the gist of the present invention. For example, in the above-described embodiment, the destination block of the motion vector in the current reference frame to be processed is predicted by referring to the motion vector in the immediately preceding reference frame. The motion vector in the frame may be referred to. This is because it is considered that there is a higher possibility that more accurate prediction can be made by referring to the motion vectors in a plurality of reference frames.

Two specific examples will be given. For example, as shown in FIG. 3B, when referring to the motion vectors (Vn-1 to Vn-5) in a plurality of continuous reference frames, it can be seen that the motion is like a rotational motion. You may understand. In this case, in the case of a rotational motion, the curvature R or the like is known, and the destination block can be predicted fairly accurately according to the motion vector Vn obtained based on the curvature.

As shown in FIG. 3C, the motion vectors (Vn-1 to Vn-6) in a plurality of continuous reference frames are set.
In some cases, it can be seen that the movement direction is reversed up and down at a predetermined cycle. In this case, if the reversal period and the moving direction are known, it is possible to predict the destination block quite accurately. For example, the predicted motion vector Vn in the case shown in FIG.
Although it cannot be determined only from the immediately preceding motion vector Vn-1, in this case, referring to the past six motion vectors,
Because of the regularity, it can be expected that the motion vector will be the same as the motion vector Vn-4 four times before.

In other words, no matter how regular the movement is, if only the motion vector in the immediately preceding reference frame is referenced, the movement of the movement block is slightly shifted or the movement direction is largely changed. In such a case, it is conceivable that the motion may shift considerably, so that the prediction of the motion cannot be determined. Therefore, it can be said that it is preferable to adopt a method of referring to motion vectors in a plurality of continuous reference frames from the viewpoint of being able to cope with various types of motion.

In the above embodiment, the DCT unit 103 is employed as the orthogonal transform means. However, as the orthogonal transform, there is another orthogonal transform other than DCT (discrete cosine transform), for example, a discrete Fourier transform (DFT). Such a conversion method may be employed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an image coding apparatus conforming to the MPEG standard including a motion vector detection apparatus according to an embodiment of the present invention.

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

FIG. 3 is an explanatory diagram relating to prediction of a destination block of a motion vector.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Image coding apparatus 101 ... Image memory 102 ... Subtraction part 103 ... DCT part 104 ... Quantization part 105 ... Variable length coding part 106 ... Transmission buffer part 107 ... Inverse quantization part 108 ... IDCT part 109 ... Addition part 110 ... frame memory unit 111 ... motion vector detection unit 112 ... motion compensation unit 114 ... image coding control unit 201 ... detection procedure selection unit 202 ... calculation execution unit 203 ... appropriate vector determination unit 204 ... detection control unit

Claims (4)

[Claims] 1. A method for use in motion-compensated inter-frame coding,
A motion vector detecting device for detecting a motion vector of a processing block corresponding to a search region of a reference frame, comprising: A prediction means for predicting a block, and a motion vector in a destination block predicted by the prediction means is preferentially detected, and the detected motion vector is used as a motion vector in a reference frame to be currently processed. A motion vector detection device, comprising: detection control means capable of executing detection. 2. The apparatus according to claim 1, wherein said detection control means is configured to be capable of performing detection in accordance with a full search procedure for individually detecting a motion vector for all search areas in said reference frame, and further according to said simple search procedure. Calculating means for calculating a predicted block of the reference frame based on the motion vector detected by performing the detection, and corresponding pixels in the predicted block of the reference frame calculated by the calculating means and the processing block of the frame currently being processed Determining a sum of absolute differences or a sum of squared differences between the two, and determining whether the calculated value is appropriate; and determining that the calculated value is not appropriate, Detection procedure instructing means for instructing the means to perform detection according to the full search procedure. Motion vector detection device according to claim 1, symptoms. 3. The determination means is not appropriate when the sum of absolute differences or the sum of squares of differences between corresponding pixels in the prediction block of the reference frame and the processing block of the frame currently being processed is equal to or greater than a predetermined reference value. The motion vector detecting device according to claim 2, wherein the determination is made. 4. A data conversion means for converting an input video signal into pixel data of a rectangular area and outputting the data, and an orthogonal transform for performing orthogonal transformation on the pixel data of the rectangular area output from the data conversion means. Means, quantizing means for dividing the orthogonally transformed data output from the orthogonal transforming means by a predetermined quantization step value and outputting the resulting data as transformed and encoded data, and changing the transformed and encoded data outputted from the quantizing means. A variable-length encoding unit that converts the output into a long code and outputs the code; an inverse quantization unit that multiplies the output from the quantization unit by the quantization step value to generate orthogonal transform data; Inverse orthogonal transformation means for performing inverse orthogonal transformation on the orthogonal transformation data output from the means and outputting pixel data in a rectangular area; and pixel data output from the inverse orthogonal transformation means and the data transformation. A motion prediction unit configured to generate difference data from the pixel data of the rectangular area output from the conversion unit and output the data to the orthogonal transformation unit; An image encoding device comprising the motion vector detection device according to any one of claims 1 to 3.