JPH07336563A - Image filtering method and image filtering device - Google Patents

Abstract

PURPOSE:To improve the image quality, to reduce information quantity of an image in the case of coding and to enhance the image quality of the coded image by eliminating noise superimposed on the image and minimizing deterioration in the image quality. CONSTITUTION:A received image signal A is given to a filter 1, from which a filtered image signal B is outputted. Then a difference is taken between the signal B and a filtered output image signal C at a preceding time having been stored in a frame memory 4 is taken and the difference is outputted to a CPU 3. A threshold level calculation means 3a of the CPU 3 calculates a threshold level signal from the difference signal D based on the received image signal A and the filtered output image signal C. Then a comparison selection means 3b compares the output of the means 3a with a threshold level with respect to each picture element and selects the filtered image signal B when the level of the signal D does not exceed the threshold level. When the threshold level is exceeded, it is discriminated that the deterioration is eminent and the received image signal A is adopted and a filtered output image signal F is generated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise elimination filter and a prefilter for encoding a digital image.

[0002]

2. Description of the Related Art In encoding a digital image on which noise is superimposed, information is allocated and encoded even to unnecessary noise. Therefore, conventionally, an image filtering method and an image filtering apparatus have been developed that minimize noise and reduce noise.

An example of the above-mentioned conventional image filtering apparatus will be described below. FIG. 4 shows an infinite impulse response filter (II
R filter). In FIG. 4, the input terminal 5 is connected to the multiplier 11. The multiplier 11 is a multiplier that multiplies the input signal by k, and the multiplier 12 is a multiplier that multiplies the input signal by 1-k, and the outputs thereof are output to the adder 13, respectively. The adder 13 adds these, and its output is given to the frame memory 4 and the output terminal 6. The frame memory 4 is a memory that holds the filtered image signal C that is previous in time, and its read signal is given to the input end of the multiplier 12. Here, A is an input image signal, C is a temporally previous filtered image signal, and B is a filtered image signal.

The operation of the conventional image filtering apparatus thus configured will be described below. The first input image signal A input from the input terminal 5 is not filtered and is set as the filtered image signal B and stored in the frame memory 4. The second and subsequent input images are filtered according to the following equation.

[Equation 1]

The input image signal A is multiplied by k in the multiplier 11.
The time-previous filter image signal C held in the frame memory 4 is multiplied by 1-k in the multiplier 12. Multiplier 11,
The outputs of 12 are added by the adder 13 to obtain the filtered image signal B. In general, since noise has no correlation in the time direction, there is no correlation between noise superimposed on the input image signal A and noise superimposed on the filtered image signal B. Therefore, in a temporally stationary image, it is possible to suppress the noise of the filtered image signal B by the smoothing action of the filter. The generated filtered image signal B is stored in the frame memory 4. After that, the same method is used for filtering.

According to the above method, noise can be sufficiently suppressed by reducing the value of k (for example,
"Digital signal processing of images", pages 97 to 98,
Nobuhiko Fukibe, Nikkan Kogyo Shimbun).

[0007]

However, in the above-described image filtering apparatus, if k is made small, the output image is likely to be deteriorated. In the case of the above-mentioned conventional example, the temporal steadiness of the image is required, and an afterimage occurs for an image that does not sufficiently satisfy the temporal steadiness condition.

Conventionally, in order to suppress the occurrence of such an afterimage, k is brought close to 1 or is set by setting k from temporally preceding and subsequent images before filtering. However, when k is brought closer to 1, the smoothing action of the filter is weakened, and the filtered image signal B becomes the input image signal A.
The effect of the filter diminishes because it approaches. On the other hand, when k is set from temporally previous and subsequent images before filtering, k must be close to 1 in a portion where an afterimage occurs so as to minimize deterioration of the image. Can't get enough. As a result, the conventional image filtering apparatus has a problem that the image quality deteriorates when the noise is sufficiently removed.

An object of the present invention is to provide an image filtering method and apparatus for solving such conventional problems, sufficiently removing noise and minimizing image quality deterioration.

[0010]

According to a first aspect of the present invention, an input image signal is filtered to generate a filtered image signal, and the filtered image signal and a temporally previous filter at a corresponding position of the image are filtered. The difference signal from the output image signal is calculated, the threshold value at that position is calculated based on the input image signal and the filter output image signal that is previous in time, the difference signal and the threshold signal are compared, and the difference signal is the threshold value. When the difference signal is smaller than the signal, the filter image signal is output to generate the filter output image signal, and when the difference signal is larger than the threshold signal, the input image signal is generated as the filter output image signal.

According to the second aspect of the present invention, the input image signal is filtered to generate a filtered image signal, and the difference between the filtered image signal and the temporally previous filtered output image signal at the corresponding position of the image. The signal is calculated, the difference signal is compared with a predetermined threshold value, and when the difference signal is smaller than the threshold value, the filter image signal is output to generate the filter output image signal, and when the difference signal is larger than the threshold value, the input image signal Is generated as a filter output image signal.

The invention according to claim 6 of the present application is characterized in that spatial filter processing is performed using the filter output image signal as an input.

According to a seventh aspect of the present invention, there are provided a filter means for filtering an input image signal, a filter image signal output from the filter means and a filter output signal temporally preceding the corresponding position of the image. Subtracting means for calculating the difference signal, threshold calculating means for calculating the threshold signal from the input image signal and the filter output image signal temporally preceding, and the threshold signal and the difference signal at each position calculated by the threshold calculating means Comparing and selecting means for selecting the filter image signal when the difference signal is smaller than the threshold signal, and selecting the input image signal and outputting as the filter output image signal when the difference signal is larger than the threshold signal, And an image memory that holds an output image signal output from the comparison / selection means.

According to the invention of claim 8 of the present application, there are provided a filter means for filtering the input image signal, a filter image signal output from the filter means and a filter output signal temporally preceding the corresponding position of the image. The subtraction means for calculating the difference signal and the predetermined threshold value and the difference signal are compared,
If the difference signal is smaller than the threshold value, the filter image signal is selected, and if the difference signal is larger than the threshold value, the input image signal is selected and output as the filter output image signal.
It is characterized by comprising a selecting means and an image memory for holding an output image signal output from the comparing / selecting means.

A twelfth aspect of the present invention is characterized by having a spatial filter means for performing a spatial filter process using the filter output image signal as an input.

[0016]

According to the filtering method and apparatus of the present invention, the filter output image signal is generated by switching between the input image signal and the filter image signal. An afterimage is generated in the filtered image signal in a portion having a large temporally unsteady motion, and the image quality is significantly deteriorated. Therefore, by adopting the input image signal in a portion where the image quality of the filtered image signal is significantly deteriorated, the image quality deterioration can be minimized.
By the above method, the noise can be sufficiently removed, the quality of the image can be improved, and the information amount of the image can be reduced at the time of encoding. Also, the quality of the coded image will be improved.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image filtering method and apparatus according to the present invention will be described below with reference to the drawings. Figure 1
FIG. 1 is an explanatory diagram showing a first embodiment of an image filtering device of the present invention. In FIG. 1, A is an input image signal and B is
Is a filter image signal, C is a filter output image signal of a previous frame, for example, one frame before, D is a difference signal, and F is a filter output image signal. The filter 1 is a filter means for filtering the input image signal A. The filter means 1 of the present embodiment is similar to the above-mentioned conventional example, and has a multiplier 11 that multiplies the input image signal A by k, a multiplier 12 that multiplies the filter output image signal C that is temporally previous by 1-k. And an adder 13 for adding the outputs of the multipliers 11 and 12 and outputting the filtered image signal B. Subtractor 2 is 1
It is subtraction means for subtracting the filter image signal B at the corresponding position from the filter output image signal C before the frame to obtain the difference signal D, the output of which is input to the CPU 3. The CPU 3 has an input image signal A and a filtered image signal B,
The filter output image C and the difference signal D output from the frame memory 4 which are temporally previous are input to generate a filter output image signal F. The function is achieved. CP
The filter output image F calculated from U3 is input to the frame memory 4. The frame memory 4 is an image memory that holds the filter output image signal F and provides the filter output image signal C. In the present embodiment, the image is composed of frames, the filtering is performed on a frame-by-frame temporal filter basis, and the threshold value calculation process and the subtraction process are performed on a pixel-by-pixel basis.

The operation of the image filtering apparatus of this embodiment having the above configuration will be described. First, when the input image signal A is input from the input terminal 5, the filter 1 performs the filtering process by the same operation as that of the conventional example described above, and the filter image signal B is generated as the output of the filter 1. The filter image signal B is input to the subtractor 2 and the difference signal D from the filter output image signal C is obtained.
The input image signal A, the filter image signal B, the filter output image signal C, and the difference signal D are input to the CPU 3. However, the input image signal A is x (t, v, h) and the filtered image signal B is y.
(t, v, h), the filter output image signal F is z (t, v, h), the differential signal D is d (t, v, h), the threshold T is T (t, v, h), and t Is the time of the frame, v is the vertical position of the pixel in the frame, and h is the horizontal position of the pixel in the frame. First, the difference signal D is shown as follows.

[Equation 2]

The current input image signal A, that is, x (t, v, h)
And the filter output image signal C one frame before, that is, z (t, v, h) held in the frame memory 4,
The threshold value calculation means 3a calculates the threshold value T (t, v, h) as in the following equation (3).

[Equation 3] Here, a and b are coefficients for weighted addition of the current input image signal A and the filter output image C one frame before,
The sum is 1. In the present embodiment, a and b are set to 0.5, but if the added value is 1, any other distribution can be used. In addition, α is a coefficient indicating the range of change when the added value thus obtained is used as a threshold value, and here, it is 0.
02.

Based on the threshold value thus obtained and the absolute value of the difference signal D from the subtractor 2, the comparison / selection means in the CPU 3 calculates the filter output signal F as in the following equation.

[Equation 4] In this way, the filter output image signal F is generated. The generated filter output image signal F is stored in the frame memory 4. In this process, the filtered image signal B is selected if the change is within the threshold T (t, v, h), and the input image signal A is selected if the change is exceeded. By performing the comparison / selection process as described above, by adopting the input image signal A in a portion where the image quality of the filtered image signal B is significantly deteriorated, it is possible to minimize the image quality deterioration and remove noise.

In the present embodiment, the filter 1 having the characteristic shown in the equation (1) is used, but the image filtering apparatus of the present invention is not limited to this, and a spatial filter, a temporal filter, a filter for performing adaptive processing, Further, various filters such as a filter obtained by combining these filters can be applied. Further, in the present embodiment, the threshold value processing shown in the formula (4) is performed, but the image filtering apparatus of the present invention is not limited to this, and a threshold value proportional to the exponential characteristic or the reciprocal of the logarithm, or comparison with a neighboring frame Various thresholds can be set, such as a threshold that is inversely proportional to the two-dimensional movement of the image obtained. In addition, although the threshold value processing is performed in pixel units in this embodiment, the image filtering apparatus of the present invention is not limited to this, and it is possible to perform processing in units of blocks, frames, or temporally a plurality of pixels. It is possible.

A second embodiment of the image filtering apparatus of the present invention will be described below. FIG. 2 is an explanatory diagram of the second embodiment of the image filtering apparatus of the present invention. In this embodiment, the same components and operations as those in the first embodiment described above are designated by the same reference numerals, and the description thereof will be omitted.

The present embodiment differs from the second embodiment in that the CPU 3 for performing threshold processing with the threshold as a constant is changed to the switch 7 and the input image signal A, the filtered image signal B and the difference signal D are input. This is the point where the filter output image signal F is generated. That is, the differential signal D is thresholded as follows using a constant threshold To.

[Equation 5]

Here, the switch 7 compares the difference signal with a predetermined threshold value, selects the filter image signal B when the difference signal is smaller than the threshold value, and selects the input image signal A when the difference signal is larger than the threshold value. Comparing / selecting means for using a filter output image signal is configured. In this way, the input image signal A can be adopted in a portion where the image quality of the filtered image signal B is significantly deteriorated to minimize the image quality deterioration and remove noise. With the above configuration, the same effect as that of the first embodiment can be obtained with a smaller hardware scale as compared with the first embodiment.

Next, a third embodiment of the present invention will be described. FIG. 3 is an explanatory diagram of the third embodiment of the present invention. In the first and second embodiments, threshold processing is performed in units of at least one pixel, but when the filter image signal B and the input image signal A are mixed in the frame of the filter output image signal F, the filter image signal B Discontinuity occurs at the boundary with the input image signal A. Therefore, the filter output image signal F may have a spatial redundancy that does not originally exist. In order to interpret this, the third embodiment adds the CPU 3 in addition to the first embodiment.
The spatial filter 8 is installed between the output terminal 6 and the output terminal 6.

In FIG. 3, L is an image signal obtained by multiplying the filter output image signal F by l, M is an image signal obtained by delaying the filter output image signal F by one pixel in the horizontal direction of the frame and multiplying it by m, and N is a filter output image. An image signal obtained by delaying the signal F by 2 pixels in the horizontal direction of the frame and multiplying by n, O is an image signal L
And the spatial filter output image signal generated by adding the image signal M and the image signal N. The output of the CPU 3 is input to the multiplier 81 and the delay devices 82 and 84 of the spatial filter 8. The multiplier 81 multiplies the filter output image signal F by 1 and outputs it as the image signal L. The delay devices 82 and 84 delay the input signal by one pixel and two pixels in the horizontal direction of the frame, respectively. The outputs of the delay devices 82 and 84 are input to the multipliers 83 and 85, respectively. The multiplier 83 is a multiplier that multiplies the image signal output from the 1-pixel delay filter by m and outputs the image signal M. The multiplier 85 receives the filter output image signal obtained by delaying the filter output image signal F by 2 pixels, multiplies this by n, and outputs the image signal N. The adder 86 adds the multipliers 81, 83 and 85 and outputs the spatial filter output image signal O.

In the present embodiment, the signal F output from the CPU 3 is further passed through the spatial filter 8 to smooth the entire frame. However, the processing of the spatial filter 8 is performed on a pixel-by-pixel basis, and the filter output image signal F is z (t, v, h),
The spatial filter output image signal O is o (t, v, h), t is the time position of the frame, v is the vertical position of the pixel in the frame, and h is the vertical position of the pixel in the frame. For the outer pixels, the pixels at the edge of the image were used. The following expression (6) shows the processing of this spatial filter. The sums of the coefficients l, m and n of the multipliers 81, 83 and 85 are set to 1, for example, l = n = 0.25 and m = 0.5.

[Equation 6]

Since the spatial filter output image O is generated by the spatial filter in this way, spatial smoothing is performed, so that the boundary between the filter image signal B and the input image signal A forming the filter output image signal F is defined. The discontinuity that occurs at is suppressed. With the above image filtering apparatus, it is possible to minimize image deterioration, remove noise, and further suppress spatially generated redundancy.

In this embodiment, the spatial filter having the characteristic shown in the equation (6) is used, but the image filtering apparatus of the present invention is not limited to this, and a filter having a different tap number or a two-dimensional spatial filter is used. Various spatial filters can be applied.

[0030]

As described above, according to the image filtering method and apparatus of the present invention, the difference between the filtered image signal and the temporally previous filtered output image signal or the temporally subsequent input image signal is set by a threshold value. By discriminating and selecting one of the input image signal and the filtered image signal, noise can be sufficiently suppressed and deterioration of image quality can be minimized. According to the second and eighth aspects, almost the same effect can be obtained with a smaller hardware scale. Claim 6
In the inventions 1 and 2, by further providing a spatial filter on the output side of these image filtering methods or devices, it is possible to suppress the discontinuity occurring at the boundary at the time of switching between the input image signal and the filtered image signal. The effect is obtained. Therefore, it is possible to further minimize image deterioration, remove noise, and suppress spatially generated redundancy.

[Brief description of drawings]

FIG. 1 is an explanatory diagram according to a first embodiment of an image filtering apparatus of the present invention.

FIG. 2 is an explanatory diagram according to a second embodiment of the image filtering apparatus of the present invention.

FIG. 3 is an explanatory diagram according to a third embodiment of the image filtering apparatus of the invention.

FIG. 4 is an explanatory diagram of a conventional image filtering method.

[Explanation of symbols]

 1 Filter 2 Subtractor 3 CPU 4 Frame Memory 5 Input Terminal 6 Output Terminal 7 Switch 8 Spatial Filter 11,12 81,83,85 Multiplier 13,86 Adder 82,84 Delay Device

Claims (12)

[Claims] 1. An input image signal is filtered to generate a filtered image signal, and a differential signal between the filtered image signal and a temporally previous filtered output image signal at a corresponding position of the image is calculated, Calculating a threshold value at that position based on the input image signal and the temporally previous filter output image signal, comparing the difference signal and the threshold signal, the difference signal is smaller than the threshold signal An image filtering method comprising outputting a filtered image signal to generate a filtered output image signal, and generating the input image signal as a filtered output image signal when the difference signal is larger than the threshold signal. 2. An input image signal is filtered to generate a filtered image signal, and a difference signal between the filtered image signal and a temporally previous filtered output image signal at a corresponding position of the image is calculated, The difference signal is compared with a predetermined threshold value, when the difference signal is smaller than the threshold value, the filter image signal is output to generate a filter output image signal, and when the difference signal is larger than the threshold value, the input image is generated. An image filtering method, wherein a signal is generated as a filter output image signal. 3. The difference signal and the threshold signal are calculated for each pixel.
Image filtering method described. 4. The image filtering method according to claim 2, wherein the difference signal is calculated for each pixel. 5. The image according to claim 1, wherein the filtering processing of the input image signal is performed by using at least one filter of a temporal filter and a spatial filter. Filtering method. 6. The spatial filter process is performed by using the filter output image signal as an input.
The image filtering method according to any one of 1. 7. Filter means for filtering an input image signal, and subtraction for calculating a difference signal between the filter image signal output from the filter means and the filter output signal at a position corresponding to the previous image in time. Means, a threshold value calculating means for calculating a threshold value signal from the filter output image signal temporally preceding the input image signal, and comparing the threshold value signal and the difference signal at each position calculated by the threshold value calculating means. A comparison / selection unit that selects a filter image signal when the difference signal is smaller than the threshold signal, and selects an input image signal when the difference signal is larger than the threshold signal to output as a filter output image signal; And an image memory that holds an output image signal output from the selection unit. 8. Filter means for filtering an input image signal, and subtraction for calculating a difference signal between the filter image signal output from the filter means and the filter output signal at the corresponding position in the image which is previous in time. And a predetermined threshold value and the difference signal are compared, and when the difference signal is smaller than the threshold value, the filter image signal is selected, and when the difference signal is larger than the threshold value, the input image signal is selected as the filter output image signal. An image filtering apparatus comprising: a comparison / selection unit for outputting; and an image memory for holding an output image signal output from the comparison / selection unit. 9. The image according to claim 7, wherein the subtraction unit calculates a difference signal for each pixel unit, and the threshold value calculation unit calculates a threshold signal for each pixel unit. Filtering device. 10. The image filtering apparatus according to claim 8, wherein the subtraction unit calculates a difference signal for each pixel. 11. The filtering process of the input image by the filter means is performed by using at least one filter of a temporal filter and a spatial filter. Image filtering device. 12. The image filtering apparatus according to claim 7, further comprising a spatial filter unit that performs a spatial filter process using the filter output image signal as an input.