JPH0670308A - Coder for motion compensation inter-frame prediction - Google Patents

Abstract

PURPOSE:To provide the motion compensation inter-frame prediction coder in which the detection precision is improved, the quantity of arithmetic operation is reduced and the propriety of a detected motion vector is discriminated in the detection of a motion vector between picture frames. CONSTITUTION:A sub block division section 116 divides an input picture and a picture of a preceding frame into plural sub blocks, motion vector detection sections 114,115 detect a motion vector of each sub block, a motion vector calculation section 112 obtains an arithmetic mean of the motion vectors and outputs the result to a frame memory and a multiplexer section. In this case, a motion vector discrimination circuit section 111 discriminates the validity of each motion vector and turns on a switch 113 when it is discriminated that a motion vector is 'valid'.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion-compensated interframe predictive coding apparatus used for video conferences, video telephones, cable televisions and the like.

[0002]

2. Description of the Related Art FIG. 7 shows the configuration of a conventional motion compensation interframe coding apparatus. The digitized input image in FIG. 7 is input to the adder 702 from the input terminal 701. The prediction value output from the frame memory 710 is also input to the adder 702, and a prediction error is output as the difference between the two. This prediction error is input to the orthogonal transformation unit 703 and subjected to orthogonal transformation. The transform coefficient is input to the quantization unit 704 and quantized. The quantized transform coefficient is input to the dequantization unit 712 and the multiplexing unit 705. Multiplexing unit 7
Reference numeral 05 multiplexes the transform coefficient quantized by the quantizer 704 and the motion vector detected by the motion vector detector 711, and outputs the multiplexed result through the terminal 706. On the other hand, the inverse quantization unit 712 inversely quantizes the quantized transform coefficient to restore the transform coefficient,
The inverse orthogonal transform unit 708 performs inverse orthogonal transform and outputs a prediction error. The prediction error obtained by the inverse orthogonal transform is the adder 7
09. The adder 709 is the frame memory 71
The prediction value from 0 and the prediction error from the inverse orthogonal transform unit 708 are added, and the reproduced pixel value is output to the frame memory 710. The frame memory 710 creates a reproduced image from the reproduced pixel value from the adder 709 and outputs it as a predicted value for the next frame. Further, the motion vector detection unit 711 is connected to the terminal 70.
A motion vector is obtained from the input image input from the first frame and the reproduced image from the frame memory 710 and output.

[0003]

However, the conventional motion compensation interframe predictive coding apparatus described above has the following three problems because it detects a motion vector by pattern matching of the entire block. (1) In order to reduce the motion vector detection accuracy to 1 pixel or less, the reference image must be interpolated, which complicates the process. (2) The calculation amount for matching increases exponentially. (3) It cannot be determined whether or not the detected motion vector matches the actual motion.

The present invention solves such a conventional problem by detecting a highly accurate motion vector with a small amount of calculation and determining whether or not the detected motion vector matches an actual motion. It is an object of the present invention to provide an excellent motion-compensated interframe predictive coding device capable of performing the following.

[0005]

According to the present invention, in a motion-compensated interframe predictive coding apparatus, a first means is a block division unit for dividing a block to be encoded into subblocks, and each divided subblock. It is characterized by comprising a motion vector detecting section for detecting a motion vector and a motion vector calculating section for calculating a motion vector of the block from the motion vector detected for each sub-block.

The second means of the present invention uses the evaluation amount α for determining whether the motion vector of the block matches the actual motion (valid / invalid) as the motion amount for each sub-block. It is characterized in that it has a minimum evaluation amount calculated at the time of vector detection, and is provided with a motion vector determination unit that determines whether the motion vector is valid or invalid based on the value of the evaluation amount α.

The third means of the present invention detects, for each sub-block, an evaluation amount β for determining whether the motion vector of the block matches the actual motion (valid / invalid). It is characterized in that a motion vector determination unit for determining whether the motion vector is valid or invalid is provided based on the value of the evaluation amount β calculated from the calculated motion vector value.

Further, the fourth means of the present invention comprises a motion vector judging section for judging the validity / invalidity of the motion vector using the two evaluation amounts α and β described in the second and third means. It is characterized by that.

[0009]

Therefore, according to the present invention, since the motion vector of the block is detected and determined from the motion vector detected from the plurality of sub-blocks having the reduced number of pixels, the accuracy is high without a significant increase in the amount of calculation. This has an effect that the motion vector can be detected and determined.

[0010]

(Embodiment 1) FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In FIG. 1, 101 is an input terminal, 102 is an adder that takes the difference between the input image and the predicted image, and outputs a prediction error signal, and 103 is an orthogonal that performs orthogonal transform on the prediction error signal output from 102 and outputs transform coefficients. A transform unit, 104 is a quantizer that quantizes the transform coefficient output from 103, 105 is a multiplexer that multiplexes and encodes the transform coefficient quantized in 104 and the motion vector obtained in 112, and 106 is an output A terminal, 107 is an inverse quantizer that inversely quantizes the transform coefficient quantized in 104, 108
Is an inverse orthogonal transform unit that inversely transforms the signal dequantized by 107, 109 is an adder that adds the prediction error signal output from 108 and the image of the frame memory 110, and 110 is one frame from the adder 109 Image is stored, the motion vector of the motion vector calculation unit 112 is used to perform motion compensation, and the predicted image is output to the adder 102.
It is also a frame memory that outputs the image of the previous frame to the sub-block dividing unit 116.

Reference numeral 111 indicates whether the motion vector is valid or invalid based on the evaluation values of the plurality of motion vectors obtained at 114 and 115.
A motion vector determination unit that determines invalidity and generates an ON / OFF control signal for the switch 113, 112 is a motion vector calculation unit that calculates a weighted average of motion vectors based on the evaluation value obtained by the motion vector detection unit, and 113 is A switch that turns “ON” when a “valid” signal is received from the motion vector determination unit 111 and outputs the motion vector of the motion vector calculation unit 112 to the frame memory 110 and the multiplexing unit 105, and 114 and 115 are motions of each sub-block. A motion vector detection unit that detects a vector and its evaluation value, 1
Reference numeral 16 is a sub-block division unit that divides the input image and the image of the previous frame into a plurality of sub-blocks.

Next, the operation of the first embodiment of the present invention will be described with reference to FIGS. 1, 2, 3 and 5.

In FIG. 1, the digital image signal input from the terminal 101 is input to the sub-block dividing unit 116 and the adder 102. The adder 102 is the frame memory 1
The difference image from the predicted image output from 10 is output to the orthogonal transformation unit 103. The orthogonal transform unit 103 converts the input difference image into, for example, a discrete cosine transform (DCT).
sine transform) and outputs the transform coefficient to the quantization unit 104. The quantization unit 104 quantizes the transform coefficient and outputs it to the inverse quantizer 107 and the multiplexing unit 105.
The multiplexing unit 105 multiplexes the motion vector detected by the motion vector calculation unit 112 with the transform coefficient quantized according to the result of the motion vector determination unit 111, encodes the result, and outputs the result through the terminal 106. Further, the inverse quantization unit 107 inversely quantizes the quantized transform coefficient, and the inverse orthogonal transform unit 108.
Output to. The inverse orthogonal transform unit 108 transforms this into the original prediction error and outputs it to the adder 109. The adder 109 outputs the prediction image output from the frame memory 110 and the prediction error image subjected to the inverse orthogonal transform to the frame memory 110 as a reproduced image. The sub-block dividing unit 116 divides the image of the previous frame output from the frame memory 110 and the image of the current frame input from the terminal 101 into two sub-blocks. Motion vector detector (1)
114 and the motion vector detection unit (2) 115 detect the motion vector for each sub-block.
An example of division into sub-blocks is shown in FIG. The motion vector detected for each sub-block is input to the vector calculation unit 112 and the motion vector determination unit 111.

The motion vector calculator 112 receives the input 2
Weighted average processing is performed by using one motion vector and the evaluation value calculated at the time of detection, and one motion vector is output.
The weighted average operation is shown in FIG. The ratio of the two evaluation values is ε1
In the following cases, the weighting coefficient A is set to 1, and when it is ε2 or more, it is set to 0. In the case of values in between, values from 0 to 1 are taken as shown in FIG. The motion vector determination unit 111 calculates the evaluation amount α from the two input motion vectors,
When this α is smaller than a threshold value ε given in advance, it is determined that the motion vector is valid, and the switch 113 is turned on. On the other hand, when α is larger than ε, it is determined that the motion vector is invalid and the switch 113 is turned off. An angle formed by two motion vectors or an inner product can be used as the evaluation value α at this time. FIG. 5 shows a block diagram of the motion vector determination unit when the angle formed by the two motion vectors is the evaluation amount α (α = θ).

Since the motion vector is detected by pattern matching between two images, there are many false detections in an image in which noise is superimposed or an image with little change, and the actual motion of the image may not match the motion vector. is there. Since such wrong motion vector becomes random, 2
The type of motion vector is different. Also, in the case of a motion vector that matches an actual motion, the directions and sizes of the two types of motion vectors are almost the same. Therefore, by using the inner product or the angular difference, which represents the difference between two motion vectors, as the evaluation amount of the motion vector, it is possible to perform the motion vector determination with high accuracy.

(Embodiment 2) In the encoder of Embodiment 1, the motion vector detected by the motion vector detecting unit (1) 114 and the motion vector detecting unit (2) 115 together with the evaluation amount β calculated at the time of motion detection Is output to the motion vector determination unit 111. At this time, a matching error when pattern matching is performed is used as the evaluation amount β. The motion vector determination unit 111 determines whether the motion vector is valid or invalid based on this evaluation amount β. FIG. 4 shows a block of this motion vector determination unit. The comparator 404 compares the two input evaluation amounts β, and the switch 401 selects the smaller one. This value is compared with the threshold value ε given in advance and the comparator 40.
In the case of comparison in 2, the motion vector is determined to be valid if ε or less, and a signal for turning on the switch 113 is output. If it is larger than the threshold value ε, it is determined to be invalid, and a signal for turning off the switch 113 is output. Thus, even if one motion vector is erroneously detected due to noise or the like, it is possible to obtain a motion vector close to the actual motion from the other motion vector.

(Embodiment 3) The motion vector determination units of the above Embodiments 1 and 2 are combined and two evaluation quantities of evaluation quantity α and evaluation quantity β are used, and the switch 113 is turned on only when the respective conditions are satisfied. Output the signal to turn on. The block at this time is shown in FIG. Since the motion vector is determined using the above-mentioned two evaluation amounts, it is possible to determine the effective / higher accuracy.
It is possible to determine invalidity.

[0018]

The present invention has the following effects, as is apparent from the above-mentioned embodiments. Detecting a motion vector for each sub-sampled sub block, calculating and calculating the motion vector of the block based on this, accurate motion vector detection without significantly increasing the calculation amount for motion vector detection, A decision can be made. Also, 2
Since the motion vector is calculated by averaging two motion vectors or a weighted average, it is possible to calculate a motion vector having a precision higher than that of the motion vector detected in each sub block.

[Brief description of drawings]

FIG. 1 is a schematic block diagram showing an embodiment of a motion compensation interframe predictive coding apparatus according to the present invention.

FIG. 2 is a diagram showing an operation of a sub-block division unit shown in FIG.

FIG. 3 is a diagram showing an operation of a motion vector calculation unit in FIG.

FIG. 4 is a diagram for explaining the operation of a motion vector determination unit when the motion vector evaluation amount of the present invention is used as a pattern matching error.

FIG. 5 is a diagram illustrating the operation of the motion vector determination unit when the motion vector evaluation amount of the present invention is an angle between vectors.

FIG. 6 is a diagram for explaining the operation of the motion vector determination unit when combining FIGS. 4 and 5;

FIG. 7 is a schematic block diagram showing a configuration method of a conventional motion compensation interframe predictive coding device.

[Explanation of symbols]

 101 Input Terminal 102 Adder 103 Orthogonal Transform Section 104 Quantization Section 105 Multiplexing Section 106 Output Terminal 107 Inverse Quantization Section 108 Inverse Orthogonal Change Section 109 Adder 110 Frame Memory 111 Motion Vector Determining Section 112 Motion Vector Calculating Section 113 Switch 114 Motion vector detection unit (1) 115 Motion vector detection unit (2) 116 Sub-block division unit 401 Switch 402 Comparator 403 Threshold value 404 Comparator 501 Angle calculation unit 502 Comparator 503 Threshold value 601 AND circuit

Claims (5)

[Claims] 1. A means for obtaining a prediction error by taking a difference between an input image and a predicted image, means for orthogonally transforming the prediction error signal, means for quantizing the orthogonally transformed prediction error signal, and the quantum. Means for dequantizing the converted signal and then inverse orthogonal transforming means, means for adding the inversely orthogonally transformed signal and the predicted image and inputting the result into a frame memory, and an output from the input image and the frame memory Means for dividing the image of the previous frame into a plurality of sub-blocks, means for detecting each motion vector from the sub-blocks, means for calculating a motion vector of the block based on the plurality of motion vectors, and the quantum Motion-compensated interframe predictive coding apparatus including a multiplexed prediction error signal and a multiplexing unit that multiplexes the motion vector and then codes and sends the coded signal. Place 2. A means for obtaining a prediction error by taking a difference between an input image and a predicted image, a means for orthogonally transforming the prediction error signal, a means for quantizing the orthogonally transformed prediction error signal, and the quantum. Means for dequantizing the converted signal and then inverse orthogonal transforming means, means for adding the inversely orthogonally transformed signal and the predicted image and inputting the result into a frame memory, and an output from the input image and the frame memory A means for dividing the image of the previous frame into a plurality of sub-blocks, a motion vector detecting means for detecting each motion vector from the sub-block and obtaining an evaluation amount thereof, and a motion vector for the block based on the plurality of motion vectors And a motion vector determination means for determining the validity of the motion vector of the block based on the plurality of motion vectors. A means for outputting to the frame memory a motion vector determined to be valid by the determination means,
A motion-compensated interframe predictive coding apparatus comprising: a multiplexing unit that multiplexes the quantized prediction error signal and the motion vector, and then encodes and sends the coded signal. 3. The motion vector determination means uses an angle between motion vectors as an evaluation amount of each motion vector, determines “effective” when the minimum value of the angles is equal to or less than a threshold value, and “other than” in the other cases. The motion-compensated interframe predictive coding apparatus according to claim 2, wherein the motion-compensated interframe predictive coding apparatus is determined to be “invalid”. 4. The motion vector determination means uses a matching error at the time of motion vector calculation as an evaluation amount of each motion vector, determines “valid” when the minimum value is equal to or less than a threshold value, and “invalid” otherwise. The motion-compensated interframe predictive coding apparatus according to claim 2, characterized in that 5. The motion vector determination means uses an angle between motion vectors and a matching error at the time of motion vector calculation as an evaluation amount of each motion vector, and the minimum value of the angle is smaller than a threshold and the matching error is minimum. The motion-compensated interframe predictive coding apparatus according to claim (2), wherein when the value is smaller than the threshold value, it is determined as “valid”, and when it is other than the above, it is determined as “invalid”.