JPS6213167A - Picture signal preprocessing system - Google Patents

Abstract

PURPOSE:To improve the coding efficiency without causing deterioration in picture quality by obtaining the difference between an original picture element value and a smoothed value of an object picture element, suppressing the differ ence value based on a predetermined characteristic and forming the processing value from the suppessed difference value and the smoothed value. CONSTITUTION:When ¦x-x''¦ is smaller than a threshold value u1, the relation of x'=x'', that is, the processing value x' to the processing object picture ele ment (x) is equal to the smoothed value x''. Random noise is reduced by the smoothing. In this case, since the absolute value of the difference between the original picture element (x) and the processed value x' has a relation of ¦x-x'¦=¦x-x''¦<u1, the picture quality deterioration due to the difference be tween the (x) and x' is suppressed to a detected limit or below by setting proper ly the threshold value u1. In case of ¦x-x''¦>u2, since the processing value is expressed as x'=x and the original picture element is preserved, the deteriora tion in the sharpness is prevented by setting properly the threshold value u2. Further, in case of u1<¦x-x''4k<u2, the processing value x' is between the x'' and the (x) and the value changes gradually from the x'' to the (x) as the value of ¦x-x''¦ increases.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical field to which the invention pertains) The present invention relates to signal processing for removing random noise contained in image signals such as television and improving correlation characteristics between pixels. This can be effectively used as preprocessing in band compression encoding of television signals.

(Prior Art) A temporal filter has been well known as a signal preprocessing method for band compression encoding of a television signal. This is a process of smoothing pixel values between two pixels located at the same spatial position on the screen, for example, in two consecutive frames.

This smoothing process using a temporal filter is effective in removing random noise in a still part of an image, but it has the following problems.

■ With temporal filters, it is generally difficult to remove noise from moving parts of images.

This is because if pixel values are smoothed between frames in a moving part, the image will become blurred, causing a large deterioration in image quality.

■ Since TV signals have large correlations between neighboring pixels within a frame, uncorrelated random noise can also be effectively removed by smoothing using neighboring pixel values within a frame, but conventional temporal filters cannot do this. Such processing is not possible.

(Objective of the Invention) An object of the present invention is to solve the problems of conventional temporal filters in removing noise from input television signals, and to make it possible to remove noise in moving parts of an image as well as to remove noise in static parts. An object of the present invention is to provide a preprocessing method that can further enhance the noise removal effect when used in combination with a temporal filter.

(Structure of the Invention) The present invention performs smoothing on a pixel to be processed using its neighboring pixels within a frame, calculates the difference between the original pixel value of the target pixel and the smoothed value, and calculates the difference between the original pixel value and the smoothed value. = The most important feature is to suppress values based on predetermined characteristics and to create processed values from suppressed difference values and smoothed values.

The basic difference from the conventional technique is that instead of smoothing pixel values between frames, smoothing is performed using neighboring pixel values within a frame.

(Example) [Example 1] FIG. 1 is a pixel arrangement diagram on a television screen in a first example of the present invention.

X represents a pixel value to be processed, a and b represent pixel values adjacent to X in the horizontal direction, and c and d represent pixel values adjacent to X in the vertical direction.

In this example, the smoothed value 7 for X is a + b 1
Next, the difference value U between X and 7 is determined. That is, u =
x - x ... (2) 3'''
1° and the difference value ° are suppressed based on predetermined characteristics to create a suppressed difference value V. That is, =4=y=F(u), where F(u) is a function of U.

Generally, F(u) is IVI≦1u1 in a small range where lul is below a certain value. When 1u1 is a certain value or more, it is set so that v=u.

FIG. 2 is a diagram showing the differential suppression characteristics in the first embodiment, and shows an example of v=F(u).

In this case, v=F(u) is given by the following equation.

・・・・・・(3) However, u2>u, >O: a determined threshold value.

The processing value X' is given by the sum of Ma and V. Therefore,
As can be seen from 2, the processed value is equal to the smoothed value in the range where the absolute value 1.1 of the difference u=x−ma between the original pixel value
x l-ma), the processed value is equal to the original pixel value in the range where 1u1 is larger than the threshold u2 (X' = x).

In addition, in the range U□< l u l < u2, the processed value takes a value between the smoothed value and the original pixel value, and 1ul=lx-
As 71 increases, it gradually changes from Ma to X.

FIG. 3 is a block diagram of the first embodiment that best represents the features of the present invention, in which 101 is an input signal terminal;
1.02 and 103 are 1 line (scanning line) time delay circuits,
104 is a smoothing operation circuit, 105 is a subtraction circuit, 106 is a difference value suppression circuit, 107 is an addition circuit, and 108 is an output terminal.

Also, or processing target pixel value, a l b l Q l
d is a neighboring pixel value of X, T is a smoothed value of X, U is a difference value between

Considering the point in time when a pixel value d is input from the input signal terminal 101, the output of the line time delay circuits 102 and 103 and the output from the tap point in the middle of 1.02 and 103 causes the pixel value a to be input as shown in the figure. , b, c, and d are obtained.

Here, a and b are pixel values one sample before and after X, respectively, and c and d are pixel values one line before and after X, respectively, and the positional relationship of these pixels on the screen is is as shown in FIG. 1 above.

In the smoothing calculation circuit 104, the average of a, b, c, d, x =
1 A direct x = -(a 10 b 10 c 1 d + x) is created.

The subtraction circuit 105 creates a difference u=x-ma between X and ma.

This difference value U is suppressed in the difference value suppression circuit 106 based on predetermined characteristics, and a suppressed difference value ■ is created. This suppression characteristic is as shown in FIG. 2 above.

In the adder circuit 107, the smoothing value Ma and the suppression difference value ■ are added to produce a processed value x'=T+V, which is output to the output terminal 108.
is output from. The value of X' is as shown in equation 7 (4) above.

As explained below, in this embodiment, the processed value X' for X is
'=7, ie, equal to the smoothed value.

This smoothing can reduce random noise. At this time, since the absolute value of the difference between the original pixel value X and the processed value X' is 1x-x'1=lx-xl<ul, the threshold value 11. By appropriately setting , image quality deterioration due to the difference between X and X' can be kept below the detection limit.

On the other hand, if 1x-xl is above a certain value, there is a high possibility that the target pixel belongs to the edge part or fine part of the image, and in such a case, the original pixel value If it is replaced, the sharpness will decrease in edge areas and fine areas, resulting in a large deterioration in image quality. In this example, l x
In the case of −x l>u2, the processed value becomes X″=X and the original pixel value is preserved, so such a decrease in sharpness can be prevented by appropriately setting the threshold value u2.

Furthermore, if ul<lx−〒1<u2, the processed value
takes a value between Y and X, and gradually changes from Ma to X as lX-71 increases.

In this case as well, since l x -x'l<u, deterioration in image quality due to the difference between X and X' can be kept below the detection limit.

In this embodiment, since the pixel value smoothing process is performed using only neighboring pixels within the frame, image quality deterioration such as blurring of moving parts of pixels essentially does not occur. Therefore, it is possible to effectively remove noise even for moving parts.

FIG. 4 is a diagram showing the effect of reducing the predicted error entropy when the first embodiment of the present invention is applied, and shows the measured value of the prediction error entropy in predictive coding. However, in this measurement, U□=u2 is used as the differential suppression characteristic.

By using the method of the present invention as preprocessing for predictive encoding, noise is effectively removed and correlation between pixels is improved, and prediction error entropy can be reduced from approximately 0.5 bits/pixel to
It can be seen that the reduction can be achieved by 1.0 bit/pixel.

Note that as a result of observing the preprocessed image in this case, it has been confirmed that no image quality deterioration occurs.

[Embodiment 2] FIG. 5 is a pixel arrangement diagram on a television screen in a second embodiment of the present invention.

The meanings of a, l), Q, d, and X are exactly the same as in FIG.

xF represents a pixel value at the same spatial position one frame before X.

The smoothing value Ma for X is the same as in the first embodiment, and the difference value u=x-7 between X and 7 is calculated.

The feature of this embodiment lies in the method of creating the suppression difference value V. That is, in this embodiment, the suppression characteristic is determined by the absolute value of the inter-frame difference, which is 1.
It is variable based on x−x, l.

At this time, the suppression difference value V is given by the following equation.

However, Fx (u) lF2 (u) is a function of U, 5
is the threshold.

FIG. 6 is a diagram showing an example of differential suppression characteristics in the second embodiment of the present invention.

The processed value X' is given by the sum of the smoothing value Ma and the suppression difference value ■, and is expressed by a linear equation.

Maju F1 (u): lx XPI≦E and 10
1 to 108 are exactly the same as in the case of FIG. 3, 109 is an input terminal for the frame difference l x -x, l, and 110 is a frame difference determination circuit.

The operation in this figure is basically the same as that in FIG. 3, so only the differences will be described.

The difference lies in that the difference suppression characteristics are adaptively changed based on the frame difference 1x-x,l.

That is, the frame difference 1 is input from the frame difference input terminal 109.
x−x,l is input, and the frame difference determination circuit 110 compares and determines with a threshold value ε, and based on the determination result, the characteristics of the difference value suppression circuit are changed.

Specifically, as shown in FIG. 6 above, 1x-xP
When 1≦E, the suppression characteristics are Fl(u), 1x-x, l
>ε, the suppression characteristic becomes F2(u).

From the above explanation, the following can be said.

(i) From Figure 6 and equation (6), Yoshiko, 1x-
xP1≦ε, that is, when the absolute value of the frame difference is less than or equal to the threshold ε, the processing value
' is equal to the smoothed value (x'-7), 1x-mal ≧u
In the range of 2, the processed value is equal to the original pixel value uN(X""
X). Moreover, in the range of ul<lx-〒1<u2, the processed value takes a value between the smoothed value and the original pixel value, and l x - x
As I increases, it gradually changes from T to x8.

(jj)l x −x, l>ε, that is, when the absolute value of the frame difference is larger than the threshold ε, the processed value X' is equal to the smoothed value (x' =x), in the range 1x-71≧u4, the processed value is equal to the original pixel value (x'=x), and in the range u, <1x-71<u4, the processed value is equal to the smoothed value and the original pixel value It takes a value between the pixel values, and gradually changes from Ma to X as the value increases by 1x-71.

(iii) As shown in Figure 6, u 1 < u 3r
112 < 114, the range of 1x-xi where the smoothing effect extends is the frame difference 1x-x, l
>ε is larger than 1x−x, l≦ε.

That is, in this embodiment, when the frame difference is large,
The smoothing effect on pixels is strengthened.

In general, if smoothing by pixels within a frame is performed too strongly, the sharpness of edge portions and fine details on the screen may decrease, resulting in deterioration of image quality.

However, since visual characteristics deteriorate in moving parts of the image, such image quality deterioration is difficult to detect.

By the way, in the moving part, the inter-frame difference LX-X,l is large.

Therefore, this method, which controls the difference suppression characteristics to strengthen the smoothing effect when the difference between frames is large, effectively utilizes the visual characteristics with respect to motion and suppresses image quality deterioration in moving parts. The removal effect can be further enhanced.

In the embodiments described above, an example was given in which a near value was used as the smoothing value, but in addition to this, weighting can be performed using an average value, an average value using more pixels, or an intermediate value of neighboring pixels. It is also possible to use

1゛'''"°1""'"hasta-pi"°""do-
We have described an example using the system difference 1x-x, l, but this
1. In addition, it is also possible to use a method that uses frame differences between pixels in the vicinity of It is.

Furthermore, in Example 2, Fl, (u
), F2 < u ), but 3
Of course, it is also possible to use the second characteristic.

(Effects of the Invention) As explained below, in the preprocessing method of the present invention, the pixel to be processed is smoothed by using neighboring pixels within the frame, so that the moving and still parts of the image are smoothed. Random noise can be removed from both to increase the correlation between pixels, and the difference 1 between the smoothed value T and the original pixel value
When x-71 is less than a certain value, the original pixel value is saved as the processed value, which has the advantage of avoiding image quality deterioration such as loss of sharpness in edges and fine details. There is.

Therefore, by using the method of the present invention as pre-processing for encoding television signals, it is possible to improve encoding efficiency without causing deterioration in image quality.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the arrangement of pixels on a television screen in the first embodiment of the present invention, FIG. 2 is a diagram showing the differential suppression characteristics in the first embodiment, and FIG. 3 best represents the features of the present invention. 4 is a block diagram of the first embodiment of the present invention. FIG. 4 is a diagram showing the effect of reducing prediction error entropy when the first embodiment of the present invention is applied. FIG. 5 is a block diagram of the second embodiment of the present invention. FIG. 6 is a diagram showing an example of the differential suppression characteristic in the second embodiment of the present invention, and FIG. 7 is a block diagram of the second embodiment of the present invention. . 101... Input signal terminal, 102.103... 1 line time delay circuit, 104
... Smoothing calculation circuit, 105 ... Subtraction circuit, 106
... Difference value suppression circuit, 107 ... Addition circuit, 108
...Output terminal, 109...Frame difference input terminal, 110...Frame difference determination circuit. X... pixel value to be processed, a, b... horizontally adjacent pixel values of X, c, d
... Pixel value adjacent to X in the vertical direction, xP... Pixel value one frame before X, T... Smoothed value of X, U
...Difference value between X and T, ■...Suppression difference value, U□1u2tLI31114...Threshold value in differential suppression characteristic, F(u), F, (u), F2(u) ・=
A function representing the differential suppression characteristic, X'...processed value. Patent applicant Nippon Telegraph and Telephone Corporation Figure 1 Figure 2

Claims (2)

[Claims] (1) For each pixel of the image signal, the value of the pixel to be processed is x, and the pixels in its vicinity are used to create a smoothed value @x@ of x, and the difference value u between x and @x@ =x−@x@ and suppress this difference value based on predetermined characteristics to create a suppression difference value, and the sum x of the smoothed value @x@ and the suppression difference value v An image signal preprocessing method characterized by creating ′=@x@+v and using this as a processed value for the original pixel value x. (2) Claim 1 ( Image signal preprocessing method described in section 1).