JP3057966B2 - Motion compensation prediction device for interlaced video - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a motion compensation prediction apparatus for coding an interlaced moving image.

[0002]

2. Description of the Related Art Moving picture communication such as a video conference and CD-RO
In a high-efficiency video coding scheme for storing video in M or the like, a frame or field screen is used.
Each screen is divided into blocks of, for example, 16 pixels × 16 lines, and high-efficiency coding is performed using intra-frame coding or inter-plane coding for coding a difference between a reference screen and a current screen by motion compensation. .

FIG. 4 shows a configuration of a general encoding apparatus.
Here, reference numeral 41 denotes a subtractor, which obtains a difference between the input screen X1 and the prediction screen X2 to generate a prediction error screen X3. 42 is an encoder such as a discrete cosine transform (DCT) or a vector quantizer, 43 is a quantizer, 44 is an inverse quantizer, and 45 is a decoding such as an inverse discrete cosine transform (IDCT) or an inverse vector quantizer. The adder 46 is an adder that adds the prediction error screen X5 and the prediction screen X2 restored by the decoder to generate a local decoding screen X6. Frame memory 4 for prediction screen
The motion detector 48 calculates the amount of motion of the input screen with respect to the reference screen stored in 7, and creates a reference screen for the obtained amount of motion with the motion compensator 49 and uses it as a prediction screen. The output of the quantizer 43 is encoded by 50 variable length encoders, multiplexed by 51 multiplexers together with data such as motion vectors obtained by 48 motion detectors, and encoded information CI
Is output as

FIG. 3 shows a motion detecting section. In this motion detection, twelve comparators are used to compare 21 prediction error signals E1 based on one-vector motion detection, 22 prediction error signals E2,..., 2n based on n-vector motion detection with two comparators. Thus, the motion detection with the smaller error signal is selected by the 13 selectors to perform motion compensation to improve the coding efficiency.

[0005]

In the above-described coding apparatus, motion detection is performed on an interlaced picture composed of a frame picture and a field picture, and a prediction method is adaptively selected on the basis of a prediction error result. It is carried out.
In this case, (1) In the case of n-vector motion detection (n = 2, 3, 4,...), N times as many motion vectors are required as in the case of one-vector motion detection. Therefore, even when the prediction error is small and the difference between the prediction efficiencies is small even if either case is used, if the prediction error in the case of n-vector motion detection is smaller, n-vector motion compensation is selected. As a result, the amount of motion vectors to be transmitted increases,
The coding efficiency decreases. (2) Even when the prediction error is large, if there is not much difference between the case of 1-vector motion detection and the case of n-vector motion detection, and if the prediction error in the case of n-vector motion detection is slightly smaller, Since detection is selected, the amount of motion vectors to be transmitted increases, and the coding efficiency decreases.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to improve the disadvantage of the above-mentioned conventional system, that is, the decrease in the coding efficiency due to the unnecessary increase of the motion vector amount.

[0007]

According to the present invention, there is provided a motion compensation prediction apparatus for interlaced moving pictures, wherein a motion vector of a block is predicted in units of blocks using an input screen and a reference screen, and motion is compensated by an error with respect to the prediction. At
Using a reference screen for the block, a plurality of motion estimating means for respectively outputting a different number of motion vectors and outputting a prediction error signal, and a prediction error value of the prediction error signal output from each motion estimating means. A preprocessing means for adding a handy value determined according to coding information including at least one of the number of motion vectors, the accumulated code amount, the quantization coefficient, and the type of the predicted image; and an output from the preprocessing means. Comparing the calculated prediction error values, and selecting a motion vector from a motion detection unit that outputs the minimum prediction error value and a selection flag indicating the minimum prediction error value. And a selection flag indicating which of the motion detection means has selected the output, and a motion vector corresponding to the selection flag. Compensation prediction apparatus is provided.

As described above, according to the present invention, one or a plurality of motion vectors are used per block in motion compensation prediction in units of blocks in an image transmission such as a high-efficiency encoder for interlaced moving images or a storage device. At this time, a handy value is given between prediction methods, and a prediction method with a small number of vectors is preferentially selected. As a result, the amount of information can be efficiently reduced, and the coding efficiency can be improved.

[0009] The handy value is determined according to the number of motion vectors, and preferably increases as the number of motion vectors increases.

When the prediction error value is smaller than a predetermined threshold value, it is preferable to preferentially select the motion compensation based on the motion vector from the motion detecting means that outputs the minimum number of motion vectors. In that case,
The motion detecting means includes a motion detecting means for outputting one motion vector and outputting a prediction error signal, and outputs one motion vector when the prediction error value is smaller than a predetermined threshold value. More preferably, the motion compensation based on the motion vector from the motion detecting means is preferentially selected. As a result, it is possible to prevent a decrease in coding efficiency due to the conventional heavy use of the field motion compensation having a large number of motion vectors, and to reduce the amount of coded information.

[0011]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In the figure, 10 is input block data, 11 is reference block data, and both are interlaced screens. Motion detection is performed on the block signals of the input screen and the reference screen by the one-vector motion detector 21, and the motion vector V1 and the prediction error E1 of the block are output. The two-vector motion detector 22 also performs motion detection, and outputs a motion vector V2 and a prediction error E2. Similarly, the n-vector motion detector 2n also performs motion detection, and outputs a motion vector Vn and a prediction error En.

The one-vector motion detector 21 outputs a motion vector obtained by predicting a motion between an input screen block and a reference screen block using one or more reference screens, and an error between the predicted screen and the input screen. I do. Similarly, the two-vector motion detector 22 predicts one screen using two or more reference screens, and the n-vector motion detector 2n predicts one screen from n or more reference screens.

The prediction errors E1, E2,..., En output from the one-vector motion detector 21, the two-vector motion detector 22,. A handy value corresponding to the number of vectors is added to each prediction error signal based on the CI. Each signal to which the handy value is added is compared by the comparator 12, and
One is selected for each block, and a selection flag ZM is output as a signal indicating which one is selected. The selector 13 selects a motion vector based on the result of the comparator 12, and outputs the selected motion vector as a motion vector ZV.

Hereinafter, a configuration of a main part of the present embodiment will be described in detail.

First, referring to FIG. 2, the configuration of block data of an input screen will be described in detail. Regarding the size of the input block, the luminance signal is 16 pixels × 16 lines, and the two chrominance signals are each 8 pixels × 16 lines. These are collectively called a macroblock. A series of encoding processes is performed for each macroblock.

As shown in the figure, the block is composed of data of an odd field present on an odd line (偶) and data of an even field present on an even line (□). The frame block is
The field block is composed of data in which odd-numbered lines and even-numbered lines appear alternately, and the field block is composed of data in which only data of odd-numbered lines are collected or data in which only data of even-numbered lines are collected.

For a reference block used for motion detection, only a luminance signal is used. The size of the block changes according to the search range. For example, if the search range is ± 7 pixels in the main and sub-scanning directions, the size is 30 pixels × 30 lines. The size of the prediction error data obtained after the motion compensation has the same configuration as that of the input block in FIG. 2 for both the luminance signal and the color difference signal.

Each of the 16-pixel × 16-line block signals input to the one-vector motion detector 21, the two-vector motion detector 22,... ,..., And n vectors. In this case, the reference screen is
The screen before or after the time can be used. Furthermore, as long as it can be represented by one, two,..., And n vectors, it is possible to refer to two or more screens before and after in time.

Next, a one-vector motion detector 21, a two-vector motion detector 22,...
FIG. 5 shows the internal configuration of n. In FIG. 5, one of the motion detectors 21, 22,..., 2n is shown as a representative. The input screen and the reference screen are input block memory 20a and reference block memory 20 respectively.
b. The input screen data and the reference screen data for performing the motion detection based on the address from the address generation circuit 55 are extracted.

Next, in the case of the one-vector motion detector 21, the reference screen data is filtered by a time / spatial filter 52 if necessary, and a prediction error calculation circuit 53
, An error between the input screen and the reference screen is calculated. In the case of the two-vector motion detector 22, the reference screen data corresponding to the two vectors is filtered by the time / spatial filter 52 and synthesized into one prediction screen. The error between the reference and the reference screen is calculated. In the case of the n-vector motion detector 2n,
The reference screen data corresponding to the n vectors is filtered by the time / spatial filter 52, combined into one prediction screen, and the prediction error calculation circuit 53 calculates the error between the input screen and the reference screen.

The optimum vector determination circuit 54 compares the prediction error signals at each search point to determine the minimum position, and calculates a motion vector V1 consisting of one, two,... Prediction error E1, motion vector V
, Motion vector Vn and prediction error E
Output n. As the prediction error signal, a cumulative sum of absolute difference values or a cumulative sum of squared difference values can be used.

The prediction error values E1, E2,..., En from the motion detectors 21, 22,..., 2n are compared by the comparator 12, and the motion detection with the smaller prediction error is selected. Output ZM. At this time, the following determination conditions are used.

(1) if (E1 <T1) then ZV
= V1 (2) else ZV = Vx Here, Vx is a motion vector giving min (E1, E2 ',..., En'), where E2 '= E2 + T2,... En + Tn.

T1 is a threshold value, and T2,
, Tn are handy values and generally T2 <T3 <... <T
n. These values correspond to coding information CI such as the number of motion vectors, the accumulated code amount, the quantization coefficient, and the type of predicted image.
Determined according to. For example, when the accumulated coding amount and the quantization coefficient are large, the threshold value T1 is set to be large, and the handy values T2, T3,... Are also set to be large, so that the coding efficiency due to the increase in the coding amount with the number of vectors is increased. To prevent the decline. Conversely, when these values are small, it is possible to take relatively small values of T1,..., Tn.

The threshold value T1 when the number of allowable motion vectors is as small as about 2 is larger than T1 when the number of allowable motion vectors is large, thereby changing the selectivity for the number of each motion vector. Thus, it is possible to set a balance point according to the number of allowable motion vectors. Further, regarding the type of the predicted image, for example, in the case of predicting from a temporally distant image, the accuracy of the motion is not so high, so that the number of allowable motion vectors is set to be small, and T1,.
It is possible to increase the value of n and reduce the code amount of the motion vector, and allocate the remaining information amount to other coding such as transform coding. Conversely, in the case of prediction from images that are close in time, since the prediction accuracy is high, the values of T1,..., Tn can be set smaller.

However, when there are a plurality of motion compensations using the same number of vectors, a handy value is given to each. For example, when the number of vectors is two and a plurality of motion compensations can be performed using a frame vector, a field vector, or the like, the motion vector Vx in Expression (2) is defined as min (E
1, E21 ', E22'..., En '). However, E21 ′ = E21 + T21, E22 ′ = E22 + T22,...... En ′ = En + Tn

Are the prediction errors of the motion detectors 21, 22,..., And T21, T22,.

In this case, a prediction error E21 obtained from a frame vector for which prediction is performed using one motion vector from each of the temporally preceding and succeeding reference screens, and two prediction errors from the temporally preceding or succeeding reference screen. The handy values T21 and T22 having different values are different from the prediction error E22 obtained from the field vector for performing prediction using the number of motion vectors.
, It is possible to give priority to either the frame vector or the field vector.

In carrying out the present invention, various modifications are possible. For example, the size of a block is 16 pixels ×
Not limited to 16 lines, various sizes such as 32 pixels × 32 lines can be applied.

For E2 ',..., En', certain coefficients C2,..., Cn are E2, such as E2 '= C2.times.E2 + T2,. , En, and then adding the handy values T2,..., Tn, it is possible to perform finer control.

[0031]

According to the present invention, a threshold value or a handy value is given to each prediction error of a plurality of motion detections, and a motion compensation prediction having a small number of vectors is prioritized for a motion prediction having a similar prediction error. In this case, the amount of coding can be reduced and the coding efficiency can be improved as compared with the conventional motion compensation method having a large number of vectors based only on the magnitude of the prediction error value.

As an example of the effect, an ISO test moving image (Flower Garden, Bicycle) is
Under the conditions of CCIR601 image format, 4M
The image quality (S / N ratio) at the time of encoding at a bit rate of bit / s is 0.5 to 1.0 d as compared with a scheme adaptively using frame motion compensation and field motion compensation.
B was improved, and the amount of motion vector information transmission was reduced by 30 to 60%.

[Brief description of the drawings]

FIG. 1 is a configuration diagram illustrating a motion detection device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of block data.

FIG. 3 is a diagram showing a conventional motion detector.

FIG. 4 is a configuration diagram of an encoding device.

FIG. 5 is a diagram illustrating a configuration of a motion vector detector.

[Explanation of symbols]

 10 Input Screen 11 Reference Screen 12 Comparator 13 Selector 14 Preprocessor 21 1 Vector Motion Detector 22 2 Vector Motion Detector 2n n Vector Motion Detector 41 Subtractor 42 Encoder 43 Quantizer 44 Dequantizer 45 Decoder 46 Adder 47 Frame Memory 48 Motion Detector 49 Motion Compensator 50 Variable Length Encoder 51 Multiplexer

Continued on the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04N ⁷ /24-7/68 H04N 11/04 G06T 7/20

Claims (4)

(57) [Claims] 1. An interlaced video motion compensation prediction apparatus for predicting a motion vector of a block in units of blocks using an input screen and a reference screen and performing motion compensation based on an error with respect to the prediction, wherein a reference screen is used for the block. A plurality of motion detecting means for respectively outputting a different number of motion vectors and outputting a prediction error signal; and a prediction error value of the prediction error signal output from each of the motion detecting means, Preprocessing means for adding a handy value determined according to coding information including at least one of a cumulative code amount, a quantization coefficient, and a type of a predicted image; and each prediction output from the preprocessing means. Comparing the error values and selecting a selection flag indicating the minimum prediction error value and a motion vector from the motion detection means that has output the minimum prediction error value. And a step of adding to the encoded prediction error a selection flag indicating which of the motion detection means has selected an output and a motion vector corresponding to the selection flag, and transmitting the selected prediction error. A motion compensation prediction device for interlaced moving images, which is a feature of the present invention. 2. The motion compensation prediction of an interlaced moving image according to claim 1, wherein the handy value is determined according to the number of motion vectors, and increases as the number of motion vectors increases. apparatus. 3. The method according to claim 2, wherein when the prediction error value is smaller than a predetermined threshold value, the motion compensation based on the motion vector from the motion detecting means that outputs the minimum number of motion vectors is preferentially selected. The motion compensation prediction apparatus for an interlaced moving image according to claim 1 or 2, wherein 4. The motion detecting means includes a motion detecting means for outputting one motion vector and outputting a prediction error signal, and when the prediction error value is smaller than a predetermined threshold value, 4. The motion compensation prediction apparatus for an interlaced moving image according to claim 3, wherein the motion compensation based on the motion vector from the motion detecting means that outputs one motion vector is preferentially selected.