JPS5917579B2 - Dynamic compensation noise reduction method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a method for reducing noise in an image signal without blurring a moving image by utilizing the correlation between an image signal delayed using a field memory, a frame memory, etc. and an original image signal. The present invention relates to a corrected noise reduction method.

Conventionally, in order to improve the signal-to-noise ratio of a television image signal whose signal-to-noise ratio has decreased, the correlation between frames in the television image signal is used to improve the signal-to-noise ratio of the television image signal, which has no correlation between frames at all. It is known to remove random noise.

A circuit for removing such noise is constructed as shown in FIG. That is, in the circuit configuration shown in Figure 1, the analog video signal from the input terminal is
The digital video signal is guided to a D converter 1 and converted into a digital video signal, and the digital video signal is written into a frame memory 3 via an adder 2.Then, the once written digital video signal is delayed by one frame period and read out. The digital video signal is guided to the D-A converter 4 and returned to the form of an analog video signal, and then taken out from the output terminal, and the digitized image signal delayed by one frame is guided to the adder O2 mentioned above to have a level ratio α. and is added to the digital video signal on the input side. Therefore, since a video signal in the form of sequentially adding digital video signals of adjacent frame periods is extracted, there are no image signal components in the image where the correlation coefficient between adjacent frames is 1. Components such as random noise that are extracted without attenuation and have no correlation between adjacent frames are averaged and attenuated, reducing the noise level and improving the signal-to-noise ratio. , as is well known. '0 However, the conventional noise reduction using the circuit configuration shown in FIG. 1, which utilizes the presence or absence of such inter-frame correlation, has a serious drawback. In other words, if there is movement in an image, the correlation between frames in that moving image portion will be low, so the signal level of that moving image portion will be reduced, similar to the noise described in 5 above. . Therefore, if the circuit configuration shown in FIG. 1 remains as it is, the image quality of the outline of the moving image will deteriorate significantly.

The following circuit configuration has been considered as a measure to prevent such image quality deterioration in the contour portion. That is, FIG. 2 shows the configuration of a conventional noise reduction circuit that prevents the deterioration of image quality in the contour portion of a moving image as described above.

In this circuit configuration, the same parts as in the circuit configuration shown in FIG. 1 are shown with the same symbols, and in order to improve the operation of the same parts as in FIG.
The digital video signal on the input side and the digital dish resolution signal delayed by one frame period from the frame memory 3 are led to a subtracter 5, the difference between the video signals between adjacent frames is extracted, and the inter-frame difference signal is obtained. The level detection circuit 6 detects the degree of difference in the video signal that occurs between frames based on the movement of the image, and depending on the level of the interframe difference signal, outputs the video signal delayed by one frame to the input side. The detection output signal of the inter-frame difference signal level is supplied to an α control circuit 7 which appropriately sets the value of the level ratio α to be added to the video signal. Therefore, in the circuit configuration shown in Figure 2, an image portion where the absolute value of the signal level of the inter-frame difference signal exceeds a certain value is determined to be a moving image portion, and a one-frame delayed video signal is fed back. The level ratio α when adding the video signal to the input video signal is set to a small value for each pixel in the moving image portion. Although the setting of the level ratio α improves the image quality deterioration of the contour portion in the circuit configuration shown in FIG. 1, the following problems still occur. One of the problems with the circuit configuration shown in the second diagram is that when the signal-to-noise ratio of the input video signal is not good, the signal level of the inter-frame difference signal generated due to the large noise occurs depending on the movement of the image. When the signal level reaches the same level as the inter-frame difference signal, it is determined that it is only a moving image part, and as mentioned above, it becomes impossible to reduce noise while preventing image quality deterioration in the contour part. Since the correction amount α is set to a small value according to the inter-frame difference signal based on noise, the noise O reduction effect is reduced even in the case of a still image.

Another problem with the circuit configuration shown in Figure 2 is that, contrary to the problem described above, in the moving image part, the correction amount α is set small according to the interframe difference signal level, so noise can be reduced at the same time. The effect of this becomes smaller accordingly, and while the outline of the moving image part is no longer blurred, the noise in that part is also hardly reduced.

It is an object of the present invention to solve the above-mentioned conventional problems and eliminate their drawbacks, and to provide a motion correction type noise reduction method that prevents image quality deterioration in the contours of moving images without reducing the noise reduction effect. It is about providing.

That is, in the noise reduction method of the present invention, the delayed output of a delay circuit having a delay time of at least one field period is positively fed back to the input side of the delay circuit at a predetermined level ratio with respect to the input image signal to reduce the input image signal. In a noise reduction method that reduces noise in the input image signal by adding the delayed output image signal to the input image signal, the delayed output image signal of the delay circuit is compared with the input both image signals to which the delayed output image signal is not added. a motion vector detection circuit that detects the direction and amount; and a motion vector correction circuit that inversely corrects and cancels out the motion of the image in the delayed output image signal according to the direction and amount of the motion of the image detected by the motion vector detection circuit. and a motion-corrected output image signal of the motion vector correction circuit is positively fed back to the input side of the delay circuit as the delayed output at the predetermined level ratio with respect to the input image signal. It is something to do.

The invention will be explained in detail below by way of example embodiments with reference to the drawings.

The third example of the configuration of the dynamic compensation noise reduction circuit according to the method of the present invention is as follows.
As shown in the figure.

In the circuit configuration shown in the third diagram, the same parts as those in the circuit configuration shown in the first diagram are labeled with the same symbols.
The difference between the circuit according to the present invention shown in the third figure and the conventional improved circuit shown in the second figure is that a motion vector based on the movement of images between adjacent frames is detected; The motion vector correction is applied to the one-frame delayed image signal fed back to the adder 2 according to the motion vector amount. That is, the one-frame delayed video signal from the frame memory 3 is transferred to the motion vector correction circuit 9.
It is returned to the adder 2 via the adder 2 and also to the subtracter 5 described above to form a difference with the input side video signal, that is, an interframe difference signal, and the interframe difference signal is sent to the improved type shown in the second figure. As in the circuit, the detection output signal is led to the level detection circuit 6 and supplied to the α suppression circuit 7 to control the set value of the correction amount α. The one-frame delayed video signal from the input side and the video signal on the input side are supplied to the motion vector detection circuit 8 to detect a motion vector based on the motion of the image, and the detected output signal is supplied to the motion vector correction circuit 9 to detect the motion vector. Vector correction is being performed. That is, by shifting the position of the pixel of the moving image in the one-frame delayed image signal in a direction in which no moving vector originally occurs in accordance with the amount of motion vector representing the movement of the image between adjacent frames, one frame delay is achieved. It is possible to easily compensate for the image quality deterioration of the contour portion that occurs due to noise reduction performed by adding the image signal to the input image signal. FIG. 4 shows an example of a schematic configuration of the dynamic level detection circuit 8 that performs this function, and FIG. 5 shows a detailed example of its configuration.

The detection circuit shown in FIG.
3-32408 (see Japanese Unexamined Patent Publication No. 54124927), the detection circuit stores the input video signal in the frame memory 10 for one frame period, thereby detecting two adjacent frames. The video signals of the two frames are simultaneously supplied to the correlation calculating section 11, and the correlation between the two frames is calculated.

In this case, the relative position of the image block in the previous frame is supplied from the shift vector generator 12. The sequential correlation values obtained between successive frames in this manner are supplied to the minimum gate 13 to obtain a motion vector output corresponding to the maximum correlation value. At the same time, the maximum correlation value is stored in the motion vector memory 14, and is supplied as an input to the shift vector generator 12 described above when calculating the correlation for the next frame. Next, in the more detailed circuit configuration of the above-mentioned motion vector detection circuit shown in FIG. However, when this motion vector detection circuit is used in the circuit configuration shown in Figure 3, the signal has already been converted into a digital video signal in the A-D converter 1. , this A-D converter 23 becomes unnecessary.

This digital video signal is guided to the current frame buffer memory 21 and the frame memory 24, and the frame memory 2
4, the negative image signal is delayed by one frame period and guided to the buffer memory 22 for the previous frame. The video signals of two adjacent frames read from the buffer memory 22 for the previous frame and the buffer memory 21 for the current frame are guided to the difference detector 20 to form a difference signal between these two frames, and the level of the difference signal is The absolute value of is determined and the value is supplied to the integrator 19. On the other hand, the addresses of the buffer memory 22 for the previous frame and the buffer memory 21 for the current frame are controlled according to the instructions stored in the instruction store 16VC, and these addresses are supplied to each buffer memory 22 and 21 via the instruction decoder 17. Then, difference detector 2
0, find the correlation for each corresponding pixel.

Thereafter, the addresses in each of the buffer memories 22 and 21 are sequentially controlled to sequentially move the pixels for which correlation is to be determined within the necessary image blocks, and the integrator 19 calculates the integral of the correlation value for each image block.

The results of the above-mentioned integration are temporarily stored, and the same operation is repeated by changing the address and performing another block shift to sequentially find results with good correlation. By repeating this operation, the address with the best correlation corresponds to the motion vector, and the motion vector is transferred to the instruction decoder 1.
7 and stored in the dynamic memory 25.

Thereafter, in the next frame, the stored contents of the dynamic memory 25 are read out and transferred to the instruction decoder 17, and the instruction decoder 1
In step 7, the address to be supplied to the previous frame buffer memory 22 is determined based on the read motion vector, and the same operation is repeated.

By repeating this operation, the motion vectors of each block are sequentially stored in the motion memory 25.

Note that instructions for performing each of the above-mentioned operations are stored in instruction store 1.
The instruction store 16 is stored in the functional circuit 1 of a minicomputer or microcomputer, for example.
Written by 5.

Further, the instruction counter 18 provides the output of the counter 18 to the address of the instruction door 16 in order to sequentially execute the above instructions. In the above-described dynamic correction noise reduction circuit according to the present invention shown in the third diagram, the two drawbacks of the conventional circuit shown in the second diagram described above are eliminated.

That is, as a result of performing the motion correction on the outline of the moving image as described above, the inter-frame correlation of the image content is high for most pixels, so as in the circuit configuration shown in the second figure, The correction amount α for noise reduction can be set to a relatively large value for all pixels without detecting the moving image portion by the level detection circuit 6. Therefore, by setting the determination level in the level detection circuit 6 to a sufficiently high value, it is possible to effectively reduce only the noise without causing image quality deterioration in the contours of the moving image. However, in view of the above, regarding the necessity or unnecessaryness of the level detection circuit 6 similar to that in the circuit configuration shown in the second figure, in actual circuit operation, the above-mentioned dynamic correction is completely impossible, regardless of the principle. Therefore, safety measures are taken to prioritize image quality deterioration of the image itself, even if the effect of noise reduction is reduced, for image parts where such motion correction malfunctions occur. Therefore, it is necessary to use the conventional level detection circuit 6 in conjunction with the conventional level detection circuit 6 for judgment in order to cope with malfunction of the dynamic correction.

Also, regarding the second runner-up in the conventional improved type mentioned above, as mentioned above, in most moving image parts, as a result of applying motion correction, the correlation between frames of image content becomes extremely strong. Therefore, even when the correction amount α for noise reduction is set to a large value, noise can be effectively reduced without causing image quality deterioration in the contours of the moving image.

Note that the following considerations need to be taken into consideration regarding the circuit operation of the above-mentioned dynamic compensation noise reduction circuit according to the present invention shown in the third diagram.

In other words, it is most preferable to detect a motion vector in the circuit configuration shown in FIG. 3 by dividing the screen into a large number of blocks and detecting a motion vector for each block. According to
This is due to differences in image quality between processes on images that have been subjected to motion correction.
There is a risk that process distortion will occur, and furthermore, there is a risk that the overall image quality will be impaired due to noise reduction.

As a measure to prevent the occurrence of such image quality deterioration, it is possible to disperse the effects of the above-mentioned block distortion on image quality by randomly shifting the blocks for motion vector detection in a pseudo-random manner for each field. It is necessary to integrate the image over time and average it over the entire screen to substantially eliminate it. Note that in an actual image, the deterioration in image quality due to the above-mentioned block distortion is most noticeable when a large area of the image is translated in parallel, so if the areas of the blocks that divide the screen are made small, this Image quality deterioration due to block distortion can be substantially eliminated. As is clear from the above description, according to the present invention, it is possible to extremely effectively perform noise reduction even on a video signal that includes many moving image portions and has a poor signal-to-noise ratio.

In other words, as a result of motion correction, the interframe correlation of image signal components other than noise becomes high even in the moving image portion, and therefore it is necessary to set the amount of correction for noise reduction to a sufficiently large value. can. Furthermore, even if the correction amount for noise reduction is set to a sufficiently large value as described above,
Since the signal component of the image content is not attenuated, the outline of the moving image will not become blurred even if noise reduction is performed effectively.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a conventional noise compensation circuit, FIG. 2 is a block diagram showing another conventional configuration, and FIG. 3 is a configuration example of a dynamic compensation noise reduction circuit according to the present invention. FIG. 4 is a block diagram showing a schematic configuration of the motion vector detection circuit in the same configuration example, and FIG. 5 is a block diagram showing a detailed configuration example of the motion vector detection circuit. be. 1...A-D converter, 2...Adder,
3...Frame memory, 4...D-A
Converter, 5...Subtractor, 6...Level detection circuit, 7...α control circuit, 8...
Motion vector detection circuit, 9... Motion vector correction circuit, 10... Frame memory, 11...
Correlation calculation unit, 12...Shift vector generation unit,
13... Minimum gate, 14... Motion vector memory, 15... Mini computer (functional circuit),
16...Rinrei store, 17...Instruction decoder, 18...Instruction operation counter, 19...
...Integrator, 20...Difference detector, 21...
... Buffer memory for current frame, 22 ... Buffer memory for previous frame, 23 ... A-D converter, 24 ... Frame memory, 25 ...
...Dynamic memory.

Claims (1)

[Claims] 1. The delay output of a delay circuit having a delay time of at least one field period is positively fed back to the input side of the delay circuit at a predetermined level ratio with respect to the input image signal, and added to the input image signal, thereby generating the input image signal. In a noise reduction method for reducing signal noise, motion vector detection detects the direction and amount of image movement by comparing the delayed output image signal of the delay circuit and the input image signal to which the delayed output image signal is not added. and a motion vector correction circuit for inversely correcting and canceling the motion of the image in the delayed output image signal according to the direction and amount of motion of the image detected by the motion vector detection circuit, the motion vector correction circuit A motion compensation noise reduction method characterized in that the motion compensation output image signal of is fed back as the delayed output to the input side of the delay circuit with the predetermined level ratio to the input image signal.