JPH0824348B2 - Contour storage filter - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a contour preserving filter that suppresses noise that does not two-dimensionally correlate in a screen in an image signal and does not cause blurring of image quality. It is a thing.

[Conventional technology]

FIG. 2 is a block diagram showing a configuration example of a conventional contour preserving filter. (1) is a 1-line delay element, (2)
Is a 1-sample delay element, (3) is a multiplier, (4) is an adder, (22) is a subtractor, (10) is a pixel sample of interest to be smoothed, (11) and (8) are screens. Left and right adjacent pixel samples on the two-dimensional array, (12) and (9) vertically adjacent pixel samples on the screen two-dimensional array, (5) absolute value calculator, and (6) comparator , (7) a selector, (19) a horizontal difference absolute value signal, (20) a vertical difference absolute value signal, (15) a horizontal smoothed horizontal filter output, (16) Is the vertical-direction smoothed vertical filter output, (21) is the horizontal-vertical smoothed horizontal-vertical filter output, (1
7) is a selector instruction signal for controlling the selector output signal, and (18) is an output pixel sample obtained by smoothing the target pixel sample.

Next, the operation will be described with reference to FIGS. 2 and 3.

Pixel samples S (m−1, n) and S (m + 1, n) adjacent to the left and right of the target pixel sample to be smoothed and pixel samples S (m, n−) adjacent to the target pixel sample above and below.
1) and S (m, n + 1), the subtracter (22) and the absolute value calculator (5) make the horizontal difference absolute value E _h and the vertical difference absolute value E _v into the following expressions, respectively. Then, the degree of pixel change is detected.

_Eh = | S (m-1, n) -S (m + 1, n) | _Ev = | S (m, n-1) -S (m, n + 1) | The absolute difference values _Eh , _Ev are It is input to the comparator (6) together with the signal obtained by multiplying each by the weighting factor α.

On the other hand, S (m, n) of the pixel sample of interest is subjected to the following calculation by a horizontal filter constituted by using horizontal adjacent samples of the adjacent samples,
It becomes the horizontal direction smoothed signal P _h (15).

P _h = {S (m−1, n) + 2 · S (m, n) + S (m + 1,
n)} / 4 Similarly, the vertical smoothing signal P is obtained by the vertical filter configured by using the adjacent samples in the vertical direction.
_v (16) is obtained, and further the horizontal and vertical direction smoothing signal P _hv is obtained by the horizontal and vertical direction filter (21) configured by using all the adjacent samples.

P _v = {S (m, n−1) + 2 · S (m, n) + S (m, n +
1)} / 4 P _hv = {S (m−1, n) + 2 · S (m, n) + S (m + 1),
n) + S (m, n−1) + 2 · S (m, n) + S (m, n + 1)} /
8 Three smoothed signals P _h , P _v , P obtained by the above calculation
_hv is input to the selector (7) and is output as the smoothing filter output signal F (18) according to the instruction of the comparator (6). The comparator (6) performs the following logical determination using the above two absolute difference values E _h and E _v , and adaptively switches the output of the selector.

When E _h αE _v ; F = P _h (output horizontal smoothed signal) E _{v When} αE _h ; F = P _v (output vertical smoothed signal) In other _cases ; F = P _hv ( That is, a horizontal / vertical smoothing signal is output) That is, when the degree of pixel change is smaller in the horizontal direction than in the vertical direction, horizontal smoothing is performed, and conversely, when the vertical direction is smaller than the horizontal direction. Vertical smoothing is performed, and horizontal and vertical smoothing is performed when horizontal and vertical are almost equal.

[Problems to be solved by the invention]

The conventional contour preserving filter is configured as described above, and the edge is detected by the variation amount of the pixel value due to the peripheral pixels of the target pixel to be smoothed, but the effectiveness and variation of the filter are detected. It does not necessarily correspond to the amount, and there was the problem that the image quality deteriorates due to erroneous determination.

The present invention has been made in order to solve the above problems, and an object of the present invention is to obtain a contour preserving filter which has a high noise suppression capability and which performs adaptive switching with few erroneous determinations so as to prevent blurring of contour lines and the like. To aim.

[Means for solving problems]

The contour preserving filter according to the present invention, for the filter adaptive switching judgment of the smoothed pixel of interest, the sum of the values of two pixels sandwiching the pixel of interest from the top, bottom, left and right, or the diagonal direction,
The signal of the difference between the target pixel value and twice the value is used.

[Action]

In the contour storage filter according to the present invention, the smoothing is performed by obtaining the signal of the difference between the sum of the values of the two pixels sandwiching the smoothed pixel of interest in the horizontal, vertical, diagonal lines and the like and the doubled value of the pixel of interest. The high-frequency components of the signal that are attenuated by the digital filter are extracted, and the switching of the filter direction is determined so that the filters are applied in the direction with the least high-frequency components.

Example of Invention

 An embodiment of the present invention will be described below with reference to the drawings.

In FIG. 1, (1) is a 1-line delay element of the image signal, (2) is a 1-sample delay element, (3) is a multiplier,
(4) is an adder, (5) is an absolute value calculator, (6) is a comparator, (7) is a selector, (8) is a pixel sample adjacent to the right of the pixel of interest to be smoothed, ( 9) is a pixel sample adjacent below the pixel of interest, (10) is a pixel sample of interest, (11) is a pixel sample adjacent to the left of the pixel of interest,
(12) is a pixel sample adjacent to the pixel of interest, (13)
Is a horizontal high frequency component signal, (14) is a vertical high frequency component signal, (15) is a horizontal filter output signal (16) is a vertical filter output signal, (17) is a selector instruction signal,
(18) is an image smoothing filter output signal.

Next, the operation will be described with reference to FIGS. 1 and 3. Each pixel sample shown by (8) to (12) in FIG. 1 is located in the relationship as shown in FIG. 3 on the two-dimensional array of the image. Let m be the horizontal pixel number, n be the vertical pixel number, and the pixel sample of interest to be smoothed be S (m, n). The target pixel sample S (m, n) is composed of four pixel samples S (m−1, n), S that are adjacent in the horizontal and vertical directions.
From (m + 1, n), S (m, n-1), S (m, n + 1), the high frequency components attenuated by the filter are calculated both horizontally and vertically. When the signal of the absolute value is A _h in the horizontal direction and A _{v in} the vertical direction, it is defined by the following equation.

_Ah = | S (m-1, n) -2.S (m, n) + S (m + 1, n) | _Av = | S (m, n-1) -2.S (m, n) + S (M, n + 1) | On the other hand, the pixel S of interest (m, n) is subjected to the following calculation by a horizontal filter configured by using horizontal adjacent samples among the adjacent samples, and smoothed in the horizontal direction. The converted signal P _h .

P _h = {S (m−1, n) + 2 · S (m, n) + S (m + 1,
n)} / 4 Similarly, the vertical smoothing signal P _v is obtained by the vertical filter configured by using the adjacent samples in the vertical direction.

P _v = {S (m, n−1) + 2 · S (m, n) + S (m, n +
1)} / 4 Two smoothed signals P _h , P _v obtained by the above calculation
Is input to the selector (7) and is output as the smoothing filter output signal F (18) according to the instruction of the comparator (6). The comparator (6) finds the smaller one of the signals A _h (13) and A _v (14) indicating the absolute value of the high frequency component attenuated by the filter, and outputs the output of the selector (7) as follows. To switch.

When A _h <A _v ; F = P _h (output horizontal smoothed signal) A _{v When} A _h ; F = P _v (output vertical smoothed signal) That is, the horizontal and vertical high frequencies By smoothing in the direction of less components, only noise that has no correlation between pixels is suppressed.

In the above-described embodiment, the horizontal and vertical two-direction smoothing circuits are provided and the high-frequency component extraction in the two directions is performed in the same manner. It goes without saying that even better image quality can be obtained by performing component extraction.

〔The invention's effect〕

As described above, according to the present invention, since the high frequency component attenuated by the filter is used for the filter direction switching judgment of the contour preserving filter, the deterioration due to the erroneous judgment of the adaptive control is reduced, and in which direction. A simple circuit can realize an image smoothing filter in which only uncorrelated noise is suppressed.

[Brief description of drawings]

FIG. 1 is a block diagram of a contour preserving filter according to an embodiment of the present invention, FIG. 2 is a block diagram showing an example of a conventional contour preserving filter, and FIG. 3 is an image signal array diagram for explaining the operation of the filter. is there. (1) is a 1-line delay element of the image signal, (2) is a 1-sample delay element, (3) is a multiplier, (4) is an adder,
(5) is an absolute value calculator, (6) is a comparator, (7) is a selector, (8) is a pixel sample adjacent to the right of the target pixel to be smoothed, and (9) is below the target pixel. Pixel sample adjacent to, (10) pixel sample of interest, (11) pixel sample adjacent to the left of the pixel of interest, (12) pixel sample adjacent to the pixel of interest, and (13) horizontal high range Component signal, (14) vertical high frequency component signal, (15) horizontal filter output signal, (16) vertical filter output signal, (17) selector instruction signal, (18) image smoothing filter. Output signal, (19) horizontal differential absolute value signal, (20) vertical differential absolute value signal, (21) horizontal and vertical smoothed horizontal and vertical filter output, (2
2) is a subtractor. In the drawings, the same reference numerals indicate the same or corresponding parts.

Claims (2)

[Claims] 1. In a discrete signal sequence obtained by scanning pixels in a raster direction, a pixel sample to be smoothed in a two-dimensional array on the screen is defined as a focused pixel sample, and the focused pixel sample is centered. A line and a sample delay element for defining a pair of horizontally adjacent samples adjacent to the left and right and a pair of vertically adjacent samples adjacent to the upper and lower sides in a two-dimensional array on the screen; Sample values adjacent to the left and right of the pixel of interest for the sample, and a horizontal smoothing filter that performs horizontal smoothing processing of the pixel of interest sample value using the pixel of interest sample value, and for the pixel of interest sample , Performing vertical smoothing of the pixel value of interest pixel by using the sample value of the pixel of interest and the sample value adjacent in the vertical direction The horizontal high-frequency component of the signal is obtained by subtracting the target pixel sample value multiplied by 2 from the sum of the sample values of the target pixel sample that are adjacent in the horizontal direction and taking the absolute value. , A multiplier, an adder, and an absolute value calculator for extracting the sum of the values of the vertically adjacent samples of the pixel of interest sample, and subtracting the pixel of interest sample value multiplied by 2 to obtain the absolute value. A multiplier, an adder, and an absolute value calculator that extract a high-frequency component in the vertical direction of the signal by comparing the high-frequency component signal input with the extracted horizontal and vertical signals, A comparator that outputs a selector instruction signal to the selector according to the size, and a filter output that is smoothed in the direction with less high frequency components according to the selector instruction signal output by the comparator is selected. Smoothing out Contour storage filter, characterized in that a selector for outputting a signal. 2. In a discrete signal sequence obtained by scanning pixels in a raster direction, a pixel sample to be smoothed in a two-dimensional array on the screen is defined as a focused pixel sample, and the focused pixel sample is centered. A pair of vertically adjacent samples that are adjacent to the left, right and above, and below in the two-dimensional array on the screen, and an upper left, and a pair of first diagonally adjacent samples that are diagonally adjacent to above right and above right. A delay element for defining a pair of second diagonally adjacent samples diagonally adjacent to the lower left, and sample values adjacent to the left and right of the pixel sample of interest,
A horizontal smoothing filter that performs a horizontal smoothing process on the target pixel sample value using the target pixel sample value; a sample value adjacent to the target pixel sample in the vertical direction of the target pixel; A vertical direction smoothing filter that performs a vertical direction smoothing process on the pixel of interest sample value using a pixel sample value; a sample value adjacent to the pixel of interest sample in a first diagonal direction of the pixel of interest; A first diagonal smoothing filter that performs a first diagonal smoothing process on the target pixel sample value using the target pixel sample value; and a second diagonal of the target pixel with respect to the target pixel sample. A second diagonal filter that performs smoothing processing on the sample pixel value of interest in the second diagonal direction by using the sample value of the pixel of interest and the sample values adjacent in the direction A multiplier for extracting the high-frequency component in the horizontal direction of the signal by subtracting the target pixel sample value multiplied by 2 from the sum of the sample values of the target pixel sample adjacent in the horizontal direction and taking the absolute value. , An adder, an absolute value calculator, and the sum of sample values of the pixel of interest sample that are adjacent in the vertical direction, the target pixel sample value multiplied by 2 is subtracted, and the absolute value is taken to obtain the vertical direction of the signal. , An adder, an absolute value calculator for extracting the high-frequency component of the target pixel sample, and subtracting the target pixel sample value multiplied by 2 from the sum of the sample values of the target pixel sample adjacent in the first diagonal direction Then, a multiplier, an adder, and an absolute value calculator that extract a high-frequency component in the first diagonal direction of the signal by taking an absolute value, and sample values that are adjacent in the second diagonal direction of the pixel sample of interest. The pixel of interest multiplied by 2 from the sum A multiplier, an adder, and an absolute value calculator that extract a high-frequency component in the second diagonal direction of the signal by subtracting the sample value and taking the absolute value; and the extracted horizontal, vertical, first diagonal , A comparator that compares the high-frequency component signals of the signals in the second diagonal direction and outputs a selector instruction signal to the selector according to the magnitude thereof, and a high comparator signal that corresponds to the selector instruction signal output by the comparator. A contour preserving filter, comprising: a selector that selects a filter output that has been smoothed in a direction having a small range component and outputs it as a smoothed output signal.