JPS621378A - Noise eliminating system for animation picture signal - Google Patents

Abstract

PURPOSE:To prevent the degradation of picture quality which occurs in a part appearing from the shade of a moving object, by using the motion of an animation picture to switch adaptively the characteristic of a nonlinear circuit. CONSTITUTION:A motor detector 2 detects the motion in accordance with a differential signal supplied to a subtracter 1. A signal indicating the motion is used to discriminate the part appearing from the shade of the moving object. The detector 2 supplies a stop signal, which switches the characteristic of a nonlinear circuit 4 or stops the operation, to the circuit 4 with respect to the part appearing from the shade of the moving object. Though the circuit 4 gives <1 gain for a minute input amplitude and gives gain 1 for a large amplitude normally, the gain is set to 1 or a value approximating 1 in dependently of the input amplitude to suppress the occurrence of the signal distortion when the stop signal is given from the detector 2. The output of an adder 5 is not only supplied to a frame memory 6 but also outputted through a lines 501 as the animation picture signal where the noise is eliminated.

Description

[Detailed description of the invention] (Industrial application field) The present invention relates to a technique for removing noise from moving image signals.

(Conventional technology and points of reference) Conventional solution 6 Utilizes the correlation between screens as shown in Figure 6! In the four-tone removal method, for example, if the difference between screens is minute, the gain is set to less than 1, for example (1/4 to 1/2).
A nonlinear circuit is used that has a nonlinear characteristic in which the gain is 1 when , is large.

It is assumed that screens 1.2 in FIG. 1 are two consecutive screens. If the object on screen 1 moves in the direction of the arrow, screen 2
Now, screen 1 is the part hidden behind the object (the shaded part)
appears.

In conventional noise removal circuits, the microscopic surface-to-surface difference generated in this shaded area is also regarded as noise, and the gain in the nonlinear circuit is reduced to less than 1, causing visible signal distortion.
This caused dirt on the screen and deteriorated the image quality. In other words, the deterioration was visible as a tail trailing behind the moving object as it moved (a huge feature of the invention). The objective is to perform noise removal that prevents image quality deterioration that occurs in the parts that appear from the shadows of the moving object by adaptively using this method.

(Structure of the Invention) According to the present invention, there is provided a noise removal method for a moving image signal using inter-frame correlation of Φ force images, which detects φ force of a moving image signal, and uses the motion detection result based on the motion detection result. A method for removing noise from a moving picture signal is obtained, which is characterized in that the background part of the moving picture signal that newly appears due to the movement L is detected, and the degree of noise removal is weakened for the detected background part.

(Principle of the Invention) In the present invention, first, a moving image signal is divided into a Φ force area and a still area. Several such branching methods are already known.

For example, in the specification of Japanese Patent Application No. 59-IC) 4110, "φ force surface i
As shown in "Elephant 1", "The Φforce Shibun Hatahiro Device", the screen is divided into roughly blocks, the absolute value of the frame difference value of each pixel in the block is added within the block, and the result of this addition is There is a method of determining the motion and staticity of the block by comparing the size with a predetermined threshold value.

1'F) Alternatively, using the method known as gradient keying, find the Φ vector for each pixel from the luminance gradient within a frame and the frame difference value, and use this Φ vector or a set of non-zero pixels to calculate the dynamic area. It can be done.

In the present invention, any method may be used to separate the motion and static. The reason why the deterioration that leaves a trail behind the Φ blade object occurs is that, as mentioned above, the nonlinear characteristics are also applied to the portion that emerges from the formation of the Φ blade object. That is,
This is because noise removal was performed indiscriminately in the shaded area in FIG. 1, and this deterioration can be prevented by switching the degree of noise removal in this area.

In this method, the Φ force detector shown in Fig. 2 detects the Φ force from the input video signal, and the time change of the Φ force detection result is used to calculate the Φ force part and βa of the moving object. A stop signal is given to the noise B♀1g to distinguish between the parts that appear, and to cut off or stop applying the characteristics of the non-linear circuit to the parts that appear from the shadow of the moving object.

The noise canceller adjusts the gain to 1 or 1 for all input signals by the stop signal given from the motion control device.
Direct (for example, 3/4.V/8, 15/16°...). In this way, there will be no deterioration in image quality.

Next, using FIG. 3, the 8 horizontal ways of the part that emerges from the shadow of the Φ blade object will be explained. Strategy 3 The shaded area A in Figure 3 indicates the position of the moving object on a certain screen. The motion detector temporarily stores this screen. The shaded part of 3rd Σ・tsuB is 3.
Figure shows the position of the body in the next screen.

Seven detectors? When screen No. 13 is input, the image signal is compared and determined between the screens with screen A stored one screen before, and a portion where a moving object changes from presence to absence is detected. This part is the part that was in the shadow of the animal body in screen A (here, it is the part that was in the shadow of the animal object. The part that appeared in is shown.

The motion detector uses the shaded area in FIG. 3C, that is, the area that emerges from the shadow of the blade object a, as a part for carefully controlling the degree of noise removal, and supplies a stop signal to the noise removal circuit.

In this manner, switching or application of the characteristics of the non-linear circuit is stopped in the portion that appears from behind the moving object, thereby suppressing the generation of quantization noise and preventing deterioration of image quality.

In this way, a noise removal method without image quality deterioration can be obtained.

(Example) An example of the present invention will be described in detail using Figure 4. An input moving picture 1 word is supplied to the subtraction group 1 via line 101.

Reduced 11 is a powerful φ power image 1 word and frame memory 6
Prediction delayed by one screen time sourced from '! 'N, and an equally divided signal is supplied as an output to the non-linear circuit 4 and also to the motion detector 2 via the line 102.

The motion detector 2 detects the motion from the difference signal supplied from the subtractor 1, uses the I word indicating the Φ force to determine the part that appears from the shadow of the Φ blade object, and detects the movement of one animal. The non-linear circuit 4 supplies a stop signal to the non-linear circuit 4 that can change the characteristics of the non-linear circuit or stop the force exerted in response to a disturbance that appears from the shadows. A gain is given to the differential A word, but the Φ force detector 2 performs a Φ force to make this gain 1 or less than 1 depending on the presence or absence of the supplied stop signal and the input signal amplitude.

The non-linear circuit 4 usually gives a gain of less than 1 for small human force amplitudes and 1 for large amplitudes, or the Φ force detector 2 gives a gain of less than 1 for large amplitudes, or the Φ force detector 2 gives a gain of less than 1 for large amplitudes. Regardless of the gain, the gain is set to 1 or close to 1 to suppress the occurrence of signal distortion.

The output of the non-linear circuit 4 is supplied to an adder 5. KaM Indigo 5
adds the output signal of the non-linear circuit 4 and the prediction signal supplied from the frame memory 6. The output of the adder 5 is supplied to the frame memory 6 and output via a line 501 as a moving image signal from which noise has been removed.

The frame memory 6 stores the moving image signal from which noise β has been removed.
After a frame time delay, compensator 1 and adder N are used as predicted signals.
Supply to vessel 5. Next, the operation of the Φ force detector 2 will be explained with reference to FIG. , the difference 1 word from compositor 1 via line 102 is (cosynthesized) to threshold 1 circuit 204.

The threshold circuit 204 determines the significance of the equally divided signals, and regards the set of significant differences as the Φ force portion.

The significant difference signal determined by the threshold circuit 204 is stored in the memory 2.
01 and the determiner 202. Memory 201
is a 1-bit frame memory that stores information representing a significant difference signal, that is, the Φ force portion.

Φ supplied from the threshold A direct circuit 204 on a certain double side A
A judger that temporarily stores a signal indicating a moving part and indicates a moving part to the stored screen when a signal indicating a moving part in the next screen B is given. 2
02 and stores a signal indicating a moving part on screen B.

The determiner 202 compares and determines whether the signal indicating the Φ-forced part of the screen A given from the memory 201 and the signal indicating the moving part of the screen B that has just been input, or whether the φ-forced part of the previous screen is φ-forced and the next one is When the screen becomes stationary, that part is determined to be a newly appeared shadow (background) part.

The 1° signal indicating a newly emerging part of the moving object determined by the determiner 202 is sent via the line 203 as a stop signal where switching or application of the characteristics of the non-linear circuit can be stopped. 4t non-linear circuits are provided.

In the above description, only the case where a motion detection method using only frame differences was used as the Φ force detection method was described, but when the gradient method is used as the motion detection method, the threshold circuit 204 is What is necessary is to replace it with a motion detection circuit that uses the gradient method.

If the stop signal is to be issued in units of blocks, a circuit for accumulating the output of the threshold circuit 204 for each block may be provided, and the output of this circuit may be configured to be supplied to the outputs of the circuit 201 and the determiner 2o2t.

(Effects of the invention) As explained in detail above, the deterioration in image quality that appears from the shadows of the moving object, where the dirt on the screen appears from time to time, has been eliminated, and the noise in other areas has been eliminated. A noise removal method that eliminates noise without deteriorating image quality is realized. As described above, when the present invention is put into practical use, the effects are extremely large.
Figures 1, 4, and 5 are diagrams for explaining an embodiment of the noise removal circuit for one bow bow system, and Figure 6 is a diagram for explaining a conventional system.

In the figure 1... Woven M-roof 2... Recommendation detector 4.
...Non-linear circuit 5...Additional M5 6...Frame memory
201... Memory 202... Determiner 204... Threshold circuit.

Mountain side 1 Middle L autopsy

Claims (1)

[Claims] A noise removal method for a moving image signal using inter-frame correlation of moving images, which detects the movement of the moving image signal and, based on the motion detection result, detects the background part of the moving image signal that newly appears due to the movement. 1. A method for removing noise from a moving image signal, characterized in that the degree of noise removal is weakened for the detected background portion.