JPH08275171A - Motion vector detector - Google Patents

Abstract

PURPOSE: To improve motion vector detection performance for forward prediction by using both 1st and 2nd motion vector detection means so as to detect a motion vector used to predict a coded image from a preceding image with high accuracy. CONSTITUTION: A 1st motion vector detection means 11 in a detection accuracy revision means 10 detects a motion vector for forward prediction predicting coding from a preceding image displayed by a motion image signal. A 2nd motion vector detection means 12 detects a motion vector for backward prediction to predict a coding image from a future image. A motion vector high accuracy detection means 13 uses both the means 11, 12 on a coding image not using the coding mode predicting the coding image from the future image to detect a motion vector to predict the coding image from the preceding image with high accuracy. An object point revision means 15 uses one of three kinds of motion vectors detected by the means 11, 12, 13 to revise an object point of the motion vector.

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 device applied to a moving image signal coding device in image recording / communication / transmission, broadcasting, etc., and more particularly to which area of an already coded screen. The present invention relates to a motion vector detection device for motion compensation predictive coding which codes a difference between motion vector information indicating whether an input image has moved and a reference area indicated by the motion vector information.

[0002]

2. Description of the Related Art The amount of information of a multi-gradation image signal is enormous, and a wide-band transmission path and a large-capacity recording medium are required for transmitting and recording an image with a normal digital signal. Therefore, a technique for compressing and encoding an image signal into a small amount of data is required in a videophone, a videoconference, a CATV, an image file device, and the like. A motion compensation predictive coding system is known as such a moving image coding system.
In this encoding method, for the partial area of the input screen, the partial area with the highest correlation is detected from the encoded screen,
The motion vector information indicating from which position on the encoded screen the partial area of the input screen has moved, and the difference between the partial area of the input screen and the partial area of the encoded screen indicated by the motion vector information are encoded. Turn into.

When encoding is performed using both the forward prediction predicted from the past screen and the backward prediction predicted from the future screen, it does not require particularly high detection capability for each of them, but is limited to only the forward prediction. In the case of encoding by using the above method, high detection capability of a motion vector for forward prediction is required. However, when encoding is limited to only the forward prediction in the related art, the motion vector detection capability of the backward prediction is shown in FIG.
It is used to expand the search range like the search range B in (a),
The search range was redundant as compared with the required search range A. The matching accuracy is as shown in FIG. 21 (c), and there is no necessary matching accuracy as shown in FIG. 21 (b). In addition, the area C of the search range A in FIG.
, The interval between search candidate points was 2 pixel accuracy, the search density was low, and optimum motion vector detection was not guaranteed.

Further, as shown in FIG. 22, if the search point in the conventional bidirectional prediction is ◯, the search center point is ⊚. In the case of the forward prediction screen, as shown in FIG. 23, the search point can be shown as ◯ with respect to the search center point ◎, and if the forward prediction screen is used, a highly accurate search density can be obtained conventionally. But you know that you can get it.

Conventionally, when coding is performed using both forward prediction and backward prediction, high detection capability is not required for each of them, but when coding is limited to only forward prediction, the above-mentioned method is used. In addition, the ability to detect a motion vector for forward prediction was insufficient.

Further, in the coding of the interlaced image with the frame structure, high detection capability is not required for each of the field type motion vector detecting means and the frame type motion vector detecting means, but the non-interlaced image is not detected. At the time of encoding, since it is limited to frame-type motion compensation, frame-type motion vector detection capability is insufficient.

In addition, MPEG2 (moving picture coding expert) is used in an interlaced image by using only forward prediction.
s group, phase 2) to efficiently code, the dual prime (du
A dedicated circuit was required as a motion vector detection means for prediction, and the scale of the device was large. Also,
Bidirectional prediction signal generation circuit and dual prime (dual pri
The scale of the device was large because each of the prediction reference signal generation circuits was a dedicated circuit.

Further, in order to perform efficient coding using only forward prediction, it is necessary to verify the possibility of using a motion vector of zero in order to reduce the code amount, and a dedicated reference for verifying that is required. Since the memory and the prediction error calculation circuit were used, the scale of the device was large.

Further, in the coding of the interlaced image in the frame structure, high detection capability is not required for each of the field type motion vector detecting means and the frame type motion vector detecting means, but the field of the interlaced image is not required. In the structure coding, the field type motion vector detection capability is insufficient because it is limited to the field type motion compensation.

[0010]

SUMMARY OF THE INVENTION According to the present invention, when only forward prediction is used, the backward prediction motion vector detecting means is operated so as to supplement the forward prediction motion vector detecting means, and the circuit scale is limited. Therefore, it is intended to solve the problem that the motion vector detection capability of the forward prediction is insufficient.

Furthermore, when encoding a non-interlaced image, the field type motion vector detecting means is operated so as to complement the frame type motion vector detecting means, and the frame type motion vector detecting ability is insufficient. The purpose is to solve the problem.

When only forward prediction is used in the coding of an interlaced image, field-type and frame-type backward prediction is used as a motion vector detecting means for dual prime prediction necessary for improving coding efficiency. It is an object of the present invention to solve the problem that the scale of the device has become large by using the motion vector detecting means of (1) and to realize the downsizing of the device while maintaining the detection performance.

It is another object of the present invention to operate the bidirectional prediction signal generation circuit as a reference signal generation circuit for dual prime prediction to solve the problem that the scale of the apparatus is large.

In the case of efficient coding using only forward prediction, one of the backward prediction motion vector detecting means is operated as a prediction error calculating means for verifying the possibility of using the zero motion vector. An object of the present invention is to solve the problem that the scale of the device is increased by eliminating the need for a dedicated reference memory and a prediction error calculation circuit for verifying it.

Further, in the case of efficient coding using only the forward prediction, one of the prediction error calculation means of the backward prediction is operated as the prediction error calculation means for verifying the possibility of using the zero motion vector. It is an object of the present invention to solve the problem that the scale of the device is left unneeded by eliminating the need for a dedicated prediction error calculation circuit for verifying it.

Further, in the coding of the interlaced image in the field structure, the frame type motion vector detecting means is operated so as to complement the field type motion vector detecting means, and the field type motion vector detecting ability is insufficient. The purpose is to solve the problem that had occurred.

[0017]

In order to solve the above problems, a motion vector detecting apparatus according to the present invention is a motion for forward prediction for predicting a coded screen from a past screen displayed by a moving image signal. First motion vector detecting means for detecting a vector; second motion vector detecting means for detecting a motion vector for backward prediction for predicting a coded screen from a future screen displayed by a moving image signal; For a coded screen that does not use a coding mode for predicting a coded screen from a future screen, both the first and second motion vector detecting means are used to predict the coded screen from the past screen. At least a detection accuracy changing means including a motion vector high accuracy detecting means for detecting a motion vector for high accuracy is provided.

Further, the motion vector detecting apparatus according to the present invention comprises first motion vector detecting means for detecting a motion vector for predicting a field structure from a past screen, and predicting a frame structure from a past screen. Second motion vector detecting means for detecting a motion vector for predicting a motion vector for predicting a field structure from a future screen, and third motion vector detecting means for detecting a motion vector for predicting a field structure from a future screen. A fourth motion vector detecting means for detecting a motion vector for predicting a frame structure is provided, and in the coding of the interlaced image,
For an encoding screen that uses the encoding mode that predicts from the immediately preceding two field screens without using the encoding mode that predicts from the future screen, the average of the past two field screens is used for prediction. Two motion vectors for
High precision motion vector detection means for detecting with high precision by the third and fourth motion vector detection means may be provided.

Further, in the motion vector detecting device according to the present invention, in a coded screen using a coding mode predicted from a past screen and a coding mode predicted from a future screen, a prediction signal from the past screen and a future signal are used. The error between the bidirectional prediction signal and the coding block due to the average of the prediction signal from the screen of is detected, and in the coding screen that does not use the coding mode to predict from the future screen, the past of the same screen position as the coding screen A prediction error amount detection means for detecting a prediction error when the frame structure is used for prediction from the screen may be provided.

Further, the motion vector detecting device according to the present invention is a first motion vector candidate detecting means for detecting a motion vector candidate for forward prediction for predicting a coded screen from a past screen displayed by an image signal. And second motion vector candidate detecting means for detecting a motion vector candidate for backward prediction for predicting an encoded screen from a future screen displayed by an image signal, and predicting an encoded screen from the future screen. In the encoding screen that does not use the encoding mode,
High-precision motion vector candidate detection means for detecting with high accuracy motion vector candidates for predicting a coded screen from the past screen using both the first and second motion vector candidate detection means, A detection accuracy changing means including
At least one of three types of motion vector candidates detected by each of the first and second motion vector candidate detection means and the high-precision motion vector candidate detection means.
A candidate score changing unit including a high precision motion vector detecting unit that changes the motion vector detection candidate score for forward prediction using a type to detect the motion vector with higher precision may be provided.

According to the present invention, both the forward predictive motion vector detecting means and the backward predictive motion vector detecting means operate as forward predictive motion vector detecting means in the forward predictive only coding.

Further, in the encoding of the non-interlaced image, both the frame type motion vector detecting means and the field type motion vector detecting means are operated as the frame type motion vector detecting means.

Further, in the case of encoding with only forward prediction, forward prediction field type motion vector detection means, backward prediction field type motion vector detection means, forward prediction frame type motion vector detection means and backward prediction. One of the frame-type motion vector detecting means is operated as the forward predicting motion vector detecting means including the motion vector zero.

When an interlaced image is encoded only by forward prediction, 1 / th of the field type of backward prediction is used.
The 2-pixel precision motion vector detection means and the backward prediction frame type 1 / 2-pixel precision motion vector detection means are operated as the reference motion vector detection means and the differential motion vector detection means in dual prime prediction.

When an interlaced image is encoded only by forward prediction, the pixel value of the reference block designated by the standard motion vector of dual prime prediction is halved by using the bidirectional prediction signal generation circuit. A difference between the data and the pixel data of the encoded block is generated, and this difference is output instead of the pixel data of the encoded block.

In the case of encoding with only forward prediction,
The bidirectional prediction error amount detecting means is used as the prediction error amount detecting means for the case where the motion vector is zero.

In the case of encoding with the field structure, both the frame type motion vector detecting means and the field type motion vector detecting means are operated as the field type motion vector detecting means.

[0028]

Since the backward prediction motion vector detection is unnecessary in the coding of the forward prediction screen, the backward prediction motion vector detection means is also used as the forward prediction motion vector detection means to search the forward prediction motion vector. The point or matching accuracy is supplemented and the motion vector detection performance of forward prediction is improved.

When encoding is performed using only the frame structure motion vector, it is not necessary to detect the field structure motion vector, and the field structure motion vector detecting means is also used as the frame structure motion vector detecting means. As a result, the motion vector search points of the frame structure are supplemented, and the motion vector detection performance of the frame structure is improved.

Further, in the case of encoding with only the forward prediction, one of the motion vector detecting means of the backward prediction is used as the prediction error calculating means for the case where the motion vector is zero, and as a reference image memory, the backward motion is performed. A prediction reference image memory can be used, and the device can be downsized.

In the coding of the forward prediction screen of the interlaced image, one of the 1 / 2-pixel precision motion vector detecting means of the backward prediction of the field structure and the frame structure is used to detect one motion vector of the dual prime prediction. Used as a means and the other is dual prime (dual
(prime) By using it as another motion vector detecting means for prediction, dual prime (dual prim)
e) The motion vector detecting means dedicated to prediction is not required, and the overall size of the device can be reduced.

In the coding of the forward prediction screen of the interlaced image, it is not necessary to generate the bidirectional prediction signal, and the addition circuit used for generating the bidirectional prediction signal is used to perform the dual prime prediction 1 By generating the difference between the data obtained by halving the pixel value of the reference block specified by one motion vector and the pixel data of the encoded block and outputting the difference instead of the pixel data of the encoded block, a dual prime (dual prime) ) It is possible to reduce the size of the entire device without requiring a dedicated prediction circuit.

In encoding the forward prediction screen, it is not necessary to detect the bidirectional prediction error amount, and the bidirectional prediction error amount detecting means is used as the prediction error amount detecting means for the motion vector zero to predict the motion vector zero. The dedicated circuit for detecting the error amount is not required, and the entire apparatus can be downsized.

When encoding with the field structure, it is not necessary to detect the frame-type motion vector detection, and the motion vector detection means for predicting the frame structure is also used as the motion vector detection means for predicting the field structure. By using this, the motion vector search point for field structure prediction is supplemented, and the motion vector detection performance for field structure prediction is improved.

[0035]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Before explaining the embodiments, the basic concept of the present invention will be explained with reference to FIG. As shown in FIG. 1, a motion vector detection device according to the present invention includes a detection accuracy changing unit 1 that changes the detection accuracy of a motion vector for predicting a coded screen.
0, a candidate point changing unit 15 that changes the candidate point of the motion vector using the motion vector detected by the detection accuracy changing unit 10, and a moving image signal is encoded based on the output of the candidate point changing unit 15. And an encoding means 17 for

The detection accuracy changing means 10 includes a first motion vector detecting means 11 for detecting a motion vector for forward prediction for predicting an encoded screen from a past screen displayed by a moving image signal, and a moving image. Second motion vector detecting means 12 for detecting a motion vector for backward prediction for predicting a coded screen from a future screen displayed by a signal.
In a coded screen that does not use a coding mode for predicting a coded screen from the future screen, both the first and second motion vector detecting means 11 and 12 are used to extract the coded screen from the past screen. Motion vector high precision detection means 1 for highly accurately detecting a motion vector for predicting a coded screen
3 and 3 are provided.

Further, the candidate point changing means 15 is one of the three types of motion vectors detected by each of the first and second motion vector detecting means 11 and 12 and the motion vector high precision detecting means 13. The third motion vector detecting means 16 for detecting the motion vector is provided in order to change the candidate point of the motion vector by using at least one.

Embodiments of the present invention will be described below with reference to the drawings. FIG. 2 is a block diagram showing the entire first embodiment of the motion vector detecting device according to the present invention. The first reference screen storage unit 102 stores the image signal input from the input terminal 101, and outputs reference screen data and macroblock data for backward prediction. FIG. 3 shows an example of this output. Output 1 is output of reference screen data for forward prediction.
Output 2 is output of macroblock data for backward prediction, output 3 is output of reference screen data for backward prediction, and B
x (x is a number indicating the input screen order) is a screen that is encoded using forward prediction and backward prediction, Ix is a screen that is encoded only by intra prediction, and Px is a screen that is encoded by other than backward prediction. .

The first motion vector detection unit 103 stores the first reference screen image for the macroblock composed of a plurality of pixels of the image signal input from the input terminal 101 and the first reference screen image storage unit 102. The motion vector from the previously input screen stored in the unit 102 is detected, and the detected motion vector is rearranged in the encoded screen order and output. FIG. 3 shows an example of rearranging the coded screen order, and the output screen order is the coded screen order. In addition,
Even in a screen that is coded only by intra prediction, a motion vector for forward prediction used for error correction is detected.

The second motion vector detection unit 104 sets the detection reference position based on the motion vector received from the first motion vector detection unit 103, and the image memory 10
For the macroblocks input by 6 in the encoding screen order as shown in FIG. 4, a motion vector with 1/2 pixel accuracy is reproduced from the past screen locally reproduced and stored in the second reference screen storage unit 105. And the prediction signal according to the motion vector is generated, and the prediction signal, the prediction mode, and the motion vector information are output.

The subtraction circuit 107 receives the prediction signal output from the second motion vector detecting section 104 and the image memory 10.
6, the difference signal from the image signal input in the encoding screen order as shown in FIG. 4 is calculated and output. The difference signal output from the subtraction circuit 107 is converted into frequency components by the orthogonal transformation unit 108 for each of the plurality of difference signals, and requantized by the quantization unit 109. The variable length coding unit 114 performs variable length coding on the requantized signal output from the quantization unit 109 together with the prediction mode and motion vector information output from the second motion vector detection unit 104, and outputs the output terminal 115. Output to. In addition, the quantization unit 109
The requantized signal output from is subjected to inverse quantization processing by the inverse quantization unit 110, and is further inversely transformed into a differential signal by the inverse orthogonal transformation unit 111. The adder circuit 113 is provided with the differential signal output from the inverse orthogonal transform unit 111 and the delay circuit 1.
A local reproduction signal is generated by addition with the prediction signal input via 12 and is stored in the second reference screen storage unit for use in generating the prediction signal of the next input screen.

FIG. 5 shows the first motion vector detecting section 1
It is a figure which shows the 2nd Example as a specific example of 03.

The forward prediction reference screen memory 204 and the backward prediction reference screen memory 205 are the first address generation circuit 2
06 and the second address generation circuit 207, the image signal input from the input terminal 201 is stored, the forward prediction reference screen memory 204 outputs the reference screen data as shown in output 1 of FIG. 3, and the backward prediction reference is performed. The screen memory 205 outputs macroblock data for backward prediction and reference screen data as shown in outputs 2 and 3 of FIG.
The macroblock data of the screen of P5, P8, and Ib of the output 2 of FIG. 3 is not the data output from the backward prediction reference screen memory 205 but the data input from the input terminal 201, and the output of FIG. The output screen of the output 3 for the screens of P5, P8, and Ib of 2 is the reference screen for forward prediction.

The forward motion vector predictor 202 detects the macroblock input from the input terminal 201.
A predetermined number of forward prediction motion vector candidates are detected from the reference screen output from the forward prediction reference screen memory 204.

The backward prediction motion vector detection unit 203 receives the control signal from the input terminal 209 and uses the backward prediction reference picture memory 205 for the picture (Bx in FIG. 3) to be coded using the backward prediction. A predetermined number of motion vector candidates for backward prediction are detected for the macroblock input from. Further, with respect to a screen (Px, Ix in FIG. 3) that is encoded without using backward prediction, the forward prediction motion vector detection unit 202 does not search for a macroblock input from the input terminal 201. The reference area of the range is read from the backward prediction reference screen memory 204, and a predetermined number of forward prediction motion vector candidates are detected. When encoding is performed without using backward prediction, the backward prediction motion vector detection unit 203 supplements the search points of the forward prediction motion vector detection unit 202 to realize a wide range of motion vector detection functions. It

Motion vector candidate selection and rearrangement section 208
Selects a predetermined number of field-type and frame-type motion vector candidates from the motion vector candidates detected by the forward motion vector predictor detection unit 202 and the backward motion vector predictor detection unit 203, and It is rearranged in the encoding screen order as shown in FIG. 4 and output to the output terminal 210.

FIG. 6 shows the second motion vector detecting section 1
It is a figure which shows the 3rd Example as a specific example of 04.

The forward prediction reference screen memory 303 and the backward prediction reference screen memory 304 are input under the control of the third address generating means 305 or the fourth address generating means 306 which operates according to the control signal input from the input terminal 302. The first motion vector detection unit 1 that stores an image serving as a reference screen among the locally reproduced images input from the terminal 301 and is input as a control signal from the input terminal 302
The read area is set according to the motion vector candidate detected in 03, and the reference image for the macroblock input from the input terminal 315 in the encoding screen order is output.

In-screen pixel interpolation means 307, 308, 30
9, 310 are the third or fourth address generating means 3
Under the control of 05 or 306, the frame-type or field-type inter-pixel interpolation is performed on the reference image output from the reference screen memory, and the pixel-outputted image is output.

When an interlaced image is encoded with a frame structure using forward prediction and backward prediction, the frame-type forward prediction 1 / 2-pixel precision motion vector detection unit 311 uses the in-screen pixel interpolation unit 307. The frame-type forward-prediction pixel-predicted interpolated image is received from the frame-type forward-prediction pixel-predicted motion vector, and the field-type forward-predicted 1 / 2-pixel precision motion vector detection unit 312 is used. A field-type forward prediction inter-pixel prediction interpolated image is received from 308, a field-type forward prediction 1 / 2-pixel precision motion vector is detected, and a field-type backward prediction 1 / 2-pixel precision motion vector detection unit 313 is described. The field-type backward prediction inter-pixel interpolation image is received from the means 307, and the A backward prediction half-pixel precision motion vector of the frame type, and a frame-type backward prediction half-pixel precision motion vector detection unit 314 receives the frame-type backward prediction inter-pixel interpolation image from the in-screen pixel interpolating means 307; Frame type backward prediction 1/2
Detect pixel-accurate motion vector.

The bidirectional prediction signal / reference block data generation unit 316 includes a forward prediction signal output from the forward prediction 1/2 pixel precision motion vector detection units 311 and 312 and a backward prediction 1/2 pixel precision motion vector detection unit. By calculating the average value of the backward prediction signals output from 314 and 313, the bidirectional prediction signals of the frame type and the field type are calculated, and the bidirectional prediction error / intra-picture prediction error detection unit 317 is output, and the input terminal 315 is input. The generated macro block data is output as it is to the frame type backward prediction ½ pixel precision motion vector detection unit 314 as reference block data.

Bidirectional prediction error / in-screen prediction error detection unit 3
Reference numeral 17 denotes an input terminal 31 for the bidirectional prediction signal input from the bidirectional prediction signal / reference block data generation unit 316.
The prediction error with respect to the macroblock data input from No. 5 and the intra-screen prediction error using the AC component of the macroblock data input from the input terminal 315 are calculated and output to the motion vector multiplexing unit 318.

The motion vector multiplexing unit 318 calculates the half
Pixel precision motion vector detection units 311, 312, 313
The motion vector and prediction error input from 314 and the prediction error input from the bidirectional prediction error / in-screen prediction error detection unit 317 are time division multiplexed and output to the mode determination unit 319.

The mode determination unit 319 determines the optimum mode and its motion vector based on the prediction error input from the motion vector multiplexing unit 318, and outputs it to the output terminal 320.

Here, in the case of coding without using backward prediction, backward prediction of the above field type is 1/2.
Pixel precision motion vector detection unit 313 and frame type backward prediction 1/2 pixel precision motion vector detection unit 314
In addition, the field type forward prediction 1/2 pixel precision motion vector detection unit 312 and the frame type forward prediction 1 /
A forward prediction field-type and frame-type forward prediction in-screen pixel interpolation image different from that supplied to the 2-pixel precision motion vector detection unit 311 is supplied, and the field-type forward prediction 1 / 2-pixel precision motion vector detection unit is supplied. Motion vector detection is performed by both the 312 and the frame type forward prediction 1/2 pixel precision motion vector detection unit 311, and the optimal vector detection capability of forward prediction is improved.

In the case of encoding without using backward prediction, the above-mentioned field type backward prediction 1/2 pixel precision motion vector detection unit 313 (or frame type backward prediction 1/2 pixel precision motion). Vector detection unit 31
In 4), a pixel prediction image in the forward prediction screen in which the motion vector becomes zero is supplied, and the motion vector is zero in a small-scale device that does not require an image memory dedicated to motion vector detection and a detection circuit. A motion vector detection for

When a non-interlaced image is encoded without using the field type prediction, the field type forward prediction 1 / 2-pixel precision motion vector detecting unit 312 and the field type backward prediction 1 are used. / 2
A frame type different from that supplied to the pixel precision motion vector detection unit 313 to the frame type forward prediction 1/2 pixel precision motion vector detection unit 311 and the frame type backward prediction 1/2 pixel precision motion vector detection unit 314. Forward predictive and backward predictive in-screen pixel interpolated images are supplied, and the frame type forward predictive 1/2 pixel precision motion vector detection unit 311 and the frame type backward predictive 1 /
Motion vector detection is performed together with the 2-pixel precision motion vector detection unit 314, and the frame-type optimum motion vector detection capability is improved.

When the interlaced image is encoded by the field structure, the frame type forward predictive 1/2 pixel precision motion vector detection unit 311 and the frame type backward predictable 1/2 pixel precision motion vector detection. The unit 314 supplies the field type forward prediction 1/2 pixel precision motion vector detection unit 312 and the field type backward prediction 1/2 pixel precision motion vector detection unit 313 with different field type forward prediction and backward prediction. The prediction intra-screen pixel interpolation image is supplied, and the field type forward prediction 1/2 pixel precision motion vector detection unit 312
Also, the motion vector detection is performed together with the field type backward prediction 1/2 pixel accuracy motion vector detection unit 313, and the field type optimal motion vector detection capability is improved.

When an interlaced image is encoded by forward prediction from the immediately preceding screen without using backward prediction, a field for detecting a 1/2 pixel precision reference vector of dual prime (dual prime). The forward prediction in-screen pixel interpolated image of the type is supplied to the backward prediction 1/2 pixel precision motion vector detection unit 313 of the field type, and the 1/2 prime precision reference vector of dual prime is detected. . Further, the modified vector of the reference vector detected by this is input to the fourth address generating means 306, and the detected reference signal of the dual prime (dual prime) is generated as the bidirectional prediction signal / reference block data generation. It is input to the section 316. The fourth address generating means 306 generates a read address for detecting a differential motion vector of dual prime and is supplied to the backward prediction reference screen memory 304. The bidirectional prediction signal / reference block data generation unit 316 calculates a difference between the data obtained by halving the pixel value of the dual prime reference prediction signal and the corresponding macroblock data, and determines the rear of the frame type. The prediction 1/2 pixel precision motion vector detection unit 314 is supplied. In-screen pixel interpolation means 31
0 generates a reference pixel for detecting a differential motion vector of dual prime, and the backward prediction ½ pixel precision motion vector detection unit 31 of the frame type is generated.
Supply to 4. The frame-type backward-prediction 1 / 2-pixel precision motion vector detection unit 314 stores the reference pixel supplied from the in-screen pixel interpolation unit 310 as 1/2 data, and the bidirectional prediction signal / reference block data generation unit 316.
Difference (or sum) from the reference block data supplied from
The amount of prediction error calculated by is used for evaluation to detect a differential motion vector of dual prime. At this time, the motion vector multiplexing unit 318 delays the motion vector and the prediction error input from the frame-type backward prediction 1 / 2-pixel precision motion vector detection unit 314 by a delay amount as compared with the case of using backward prediction. The amount is reduced by one macroblock period, time division multiplexing is performed, and the result is output to the mode determination unit 319. Thus, 1/2 for backward prediction
Pixel precision motion vector detection means dual prime (du
al prime) reference motion vector and differential motion vector detection.
Therefore, it is possible to reduce the size of the apparatus while maintaining the detection performance of the reference motion vector and the differential motion vector of the dual prime (dual prime) without requiring a dedicated motion vector detecting means. Further, also in the bidirectional prediction signal / reference block data generation unit 316, by using the addition circuit for bidirectional prediction signal generation as the subtraction circuit for reference block generation, it is possible to downsize the device by sharing the circuit. Become.

FIG. 7 shows the second motion vector detecting section 1 described above.
It is a figure which shows the 4th Example as another specific example of 04.
Here, in particular, the reference screen memory 321 for zero motion vector
For a coded screen that does not use backward prediction, the prediction error calculation circuit included in the bidirectional prediction error / intra-prediction error detection unit 322 uses a zero motion vector instead of the bidirectional prediction error. Calculate the prediction error. Therefore, the backward prediction motion vector detecting unit (31
3 and 314) can be used for the motion vector detection of the forward prediction, and a dedicated prediction error calculation circuit for the zero motion vector is unnecessary, and the device can be downsized.

FIG. 8 is a block diagram showing a fifth embodiment of the motion vector detecting device according to the present invention. In FIG. 8, the motion vector detected for the even field of the block to be coded is detected by the block 22 on the basis of the motion vector detected for the odd field of the block to be coded. A field motion vector with 1/2 pixel precision is detected with reference to the two optimum field motion vectors detected one for each of the odd field and the even field of the block to be encoded. Of the two field motion vectors, the field motion vector with the smallest prediction error
/ 2 pixel precision frame motion vector is detected.

FIG. 9 is a block diagram showing a sixth embodiment of the motion vector detecting device according to the present invention. The sixth embodiment is intended to improve the search accuracy (matching accuracy) in the case of a non-interlaced image.
The blocks 21 and 22 calculate the field prediction error from different regions for the odd and even fields of the block to be encoded. A field prediction error that is a frame prediction error is calculated together with the field prediction error calculated by the blocks 21 and 22, and a motion vector that minimizes the frame prediction error is detected from each search region in block 23. Among the four frame motion vectors detected from each search area, a block 24 detects a 1/2 pixel precision frame motion vector with reference to a plurality of frame motion vectors having small prediction errors.
Even if the prediction error is large for the 1-pixel precision frame motion vector, there is a case where the prediction error is very small for the 1 / 2-pixel precision frame motion vector in the vicinity thereof. Multiple 1
By using the pixel precision motion vector as a reference,
There is a high possibility that 2-pixel precision motion vector detection can be performed.

FIG. 10 is a block diagram showing a seventh embodiment of the motion vector detecting device according to the present invention. This 7th
The embodiment is intended to improve search accuracy (matching accuracy) in the case of an interlaced image.
Based on the motion vector detected by the block 21 for the odd field of the block to be encoded, the block 22 detects the motion vector for the even field of the block to be encoded.

A field prediction error that is a frame prediction error is calculated together with the field prediction error calculated by the block 24, and a frame motion vector that minimizes the frame prediction error is detected from each search area. A block 25 detects a 1/2 pixel precision frame motion vector based on the field motion vector having a small prediction error among the two frame motion vectors detected from each area.

Since the 1 / 2-pixel precision frame motion vector is detected with reference to the 1-pixel precision frame motion vector, even if the reference motion vector is limited to one, the optimum 1
There is a high possibility that a / 2 pixel precision frame motion vector can be detected.

FIG. 11 is a block diagram showing a motion vector detecting device according to the eighth embodiment of the present invention. This 8th
The embodiment is intended to improve the search density in the case of a non-interlaced image. In FIG. 11, for the odd field and the even field of the block to be encoded, the block 21 and the block 26 detect the field motion vector from different regions. From the search points thinned out in blocks 21 and 26, field motion vectors are detected by blocks 22 and 27. A block 24 detects a 1/2 pixel precision motion vector based on the four field motion vectors detected from each area. This makes the search intervals uniform and increases the possibility that an optimal frame motion vector will be detected.

FIG. 12 is a block diagram showing a motion vector detecting device according to the ninth embodiment of the present invention. This 9th
The embodiment is applied to the case of a non-interlaced image, and determines the coding mode of a bidirectional prediction screen (B-picture). In FIG. 12, the frame motion vector is detected based on two optimum field motion vectors detected for each of the odd field and the even field of the block to be encoded. The two optimum field motion vectors are detected for the odd field and the even field of the block to be coded, and the reference signals are different. It is not possible to guarantee the determination of the optimum field motion vector for the vector detection standard. This increases the possibility that an optimal 1/2 pixel precision frame motion vector will be detected.

FIG. 13 shows a motion vector detecting device according to a tenth embodiment of the present invention. This tenth embodiment is different from the invention of claim 3 in that in the case of an interlaced image, a "forward prediction screen ( P-picture) Interval> 1 ”, the coding mode of the forward prediction screen is determined. The operation of each block corresponds to that of the fifth to ninth embodiments, and thus redundant description will be omitted.

FIG. 14 shows a motion vector detecting device according to an eleventh embodiment of the present invention. This eleventh embodiment also corresponds to the invention of claim 3 and the forward prediction screen (P- picture) encoding mode is determined. The operation of each block is the fifth to ninth.
Since this corresponds to the embodiment, duplicated explanation is omitted.

FIG. 15 shows a motion vector detecting device according to a twelfth embodiment of the present invention. This twelfth embodiment also corresponds to the invention of claim 2, and in the case of an interlaced image, "forward prediction" is performed. The coding mode of the forward-predicted screen is determined by "screen (P-picture) interval = 1". The operation of each block corresponds to that of the fifth to ninth embodiments, and thus redundant description will be omitted.

FIG. 16 shows a motion vector detecting device according to a thirteenth embodiment of the present invention. This thirteenth embodiment also corresponds to the invention of claim 2 and the "forward prediction screen ( P-picture) interval = 1 "is used to determine the coding mode of the forward prediction screen. The operation of each block corresponds to that of the fifth to ninth embodiments, and thus redundant description will be omitted.

FIG. 17 is an explanatory diagram showing the matching of the search points of the motion vector detected by each of the above embodiments, FIG. 17 (a) shows the matching size of the prediction of the frame structure, and FIG. 17 (b). Shows the matching size of the prediction of the field structure, and FIG. 17C shows the matching point due to the improvement of the search accuracy of the motion vector detection with the 1-pixel accuracy in the case of the forward prediction screen. FIG. 17 (c) is for detecting two motion vectors by predicting the field structure and using them for motion vector detection by predicting the frame structure. 17 (a) (b)
In FIG. 17C, white circles indicate matching points of odd field predicted field structures, and black circles in FIG. 17C indicate matching points of even field predicted field structures.

FIG. 18 shows the forward and backward bidirectional prediction 1
This is for explaining the motion vector detection with pixel accuracy. In the figure, the double circle (◎) is the search center point and the square (□) is 2.
Time search points, white circles (O) indicate search points for odd fields by field structure prediction, and black circles indicate search points for even field by field structure prediction.

FIG. 19 is a diagram for explaining motion vector detection with one pixel accuracy in the case of forward prediction. In the figure,
White circles (◯) indicate search points by prediction in the frame structure, double circles (⊚) indicate search center points, and squares (□) indicate double search points.

FIG. 20 shows pixels for motion vector detection with 1-pixel precision in the case of the forward prediction screen.
Double circles (⊚) indicate search center points, white circles (∘) indicate search points for odd fields by field structure prediction, and black circles indicate search points for even field by field structure prediction.

[0076]

According to the present invention, by effectively utilizing the detection function and arithmetic function provided, it is possible to always perform high-performance motion vector detection for any coding mode with a small-scale device. Becomes

[Brief description of drawings]

FIG. 1 is a block diagram showing the basic concept of a motion vector detection device according to the present invention.

FIG. 2 is a block diagram showing an overall configuration common to the embodiments of the present invention.

FIG. 3 is a diagram for explaining a first embodiment of a first motion vector detection section.

FIG. 4 is a diagram showing an example of rearranging an encoding screen order.

FIG. 5 is a diagram showing a second embodiment of the first motion vector detection section.

FIG. 6 is a diagram showing a third embodiment of the second motion vector detection section.

FIG. 7 is a diagram showing a fourth embodiment of the second motion vector detection section.

FIG. 8 is a block diagram showing another motion vector detecting apparatus according to a fifth embodiment of the present invention.

FIG. 9 is a block diagram showing another motion vector detecting device according to a sixth embodiment of the present invention.

FIG. 10 is a block diagram showing another motion vector detecting device according to a seventh embodiment of the present invention.

FIG. 11 is a block diagram showing another motion vector detecting device according to an eighth embodiment of the present invention.

FIG. 12 is a block diagram showing another motion vector detecting device according to a ninth embodiment of the present invention.

FIG. 13 is a block diagram showing another motion vector detecting device according to a tenth embodiment of the present invention.

FIG. 14 is a block diagram showing another motion vector detecting device according to an eleventh embodiment of the present invention.

FIG. 15 is a block diagram showing another motion vector detecting device according to a twelfth embodiment of the present invention.

FIG. 16 is a block diagram showing another motion vector detecting device according to a thirteenth embodiment of the present invention.

FIG. 17 is a diagram illustrating a pixel matching state on a screen in which a motion vector is detected according to the present invention.

FIG. 18 is a diagram showing motion vector detection with 1-pixel accuracy in the case of bidirectional prediction according to the present invention.

FIG. 19 is a diagram showing motion vector detection with 1-pixel accuracy in the case of forward prediction according to the present invention.

FIG. 20 is a diagram showing motion vector detection with 1-pixel accuracy on a forward prediction screen according to the present invention.

FIG. 21 is a diagram showing a conventional motion vector detection screen and a pixel, respectively.

FIG. 22 is a diagram showing search points and search center points in the case of conventional bidirectional prediction.

FIG. 23 is a diagram showing a search point and a search center point in the case of conventional forward prediction.

[Explanation of symbols]

10 detection accuracy changing means 11 first motion vector detecting means 12 second motion vector detecting means 13 motion vector high accuracy detecting means 15 candidate point changing means 16 (third) motion vector detecting means 17 encoding means 102 reference image Memory 103 First motion vector detection unit 104 Second motion vector detection unit 105 Local reproduction image memory 202 Forward prediction motion vector detection unit 203 Backward prediction motion vector detection unit 208 Motion vector candidate selection rearrangement unit 204, 303 Forward prediction reference Screen memory 205, 304 Backward prediction reference screen memory 311 Frame type forward prediction 1/2 pixel precision motion vector detection unit 314 Frame type backward prediction 1/2 pixel precision motion vector detection unit 312 Field type forward prediction 1/2 pixel Precision motion vector detector 13 field types backward prediction half-pixel precision motion vector detector 316 bidirectional prediction signal and the reference block detection unit 321 zero motion vector for the reference picture memory generating unit 317,322 bidirectional prediction error and intra prediction error data

Claims (4)

[Claims] 1. A first motion vector detecting means for detecting a motion vector for forward prediction for predicting a coded screen from a past screen displayed by a moving image signal, and a future motion vector displayed by a moving image signal. Second motion vector detecting means for detecting a motion vector for backward prediction for predicting an encoded screen from the screen, and an encoded screen not using an encoding mode for predicting the encoded screen from the future screen, A high-precision motion vector detecting means for detecting a motion vector for predicting a coded screen from the past screen with high accuracy by using both the first and second motion vector detecting means, A motion vector detecting device including at least a changing unit. 2. A first motion vector detecting means for detecting a motion vector for predicting a field structure from a past screen, and a motion vector for predicting a frame structure from a past screen. A second motion vector detecting means, a third motion vector detecting means for detecting a motion vector for predicting a field structure from a future screen, and a motion vector for predicting a frame structure from a future screen. A fourth motion vector detecting unit for detecting is provided, and in encoding an interlaced image, prediction is performed from the immediately preceding two field screens without using an encoding mode that is predicted from a future screen. In a coded screen using the coding mode, two motion vectors for predicting the average of the past two field screens are used as the third and third motion vectors. Motion vector detecting apparatus comprising: a, a highly accurate motion vector detection means for detecting with high accuracy in four of the motion vector detection means. 3. In a coded screen using a coding mode for predicting from a past screen and a coding mode for predicting from a future screen, both directions based on an average of a prediction signal from the past screen and a prediction signal from the future screen. Prediction when detecting the error between the prediction signal and the coding block and predicting from the future screen In the coding screen that does not use the coding mode, prediction when predicting with the frame structure from the past screen at the same screen position as the coding screen A motion vector detecting device comprising a prediction error amount detecting means for detecting an error. 4. A first motion vector candidate detecting unit for detecting a motion vector candidate for forward prediction for predicting a coded screen from a past screen displayed by an image signal, and a future motion vector candidate detecting unit displayed by an image signal. Second, detecting motion vector candidates for backward prediction that predicts an encoded screen from the screen
Of the motion vector candidate detecting means and the coding screen which does not use the coding mode for predicting the coding screen from the future screen, both the first and second motion vector candidate detecting means are used to A detection accuracy changing unit including a high-precision motion vector candidate detection unit that detects a motion vector candidate for predicting a coded screen from a past screen with high accuracy; and the first and second motion vector candidate detection units. And at least one of the three types of motion vector candidates detected by each of the high precision motion vector candidate detecting means.
A motion vector detection device comprising: a candidate point number changing unit that includes a high-precision motion vector detection unit that changes the motion vector detection candidate number for forward prediction using a type to detect a motion vector with higher accuracy.