JPS62208764A - Noise reduction device for television signal - Google Patents

Abstract

PURPOSE:To effectively reduce noise without deterioration of picture quality by distinguishing the noise and the picture signal spatially but also timewise. CONSTITUTION:A present input television signal (x) inputted from a terminal 1 is subtracted fro a forecast signal x' from a forecast circuit 41 by a subtractor 11 and a forecast error signal epsilon is obtained. Only a high frequency component epsilonH is extracted from the forecast error signal at a high pass filter 51, converted into a correction signal EH' at a correction signal generating circuit 31, added to the present input television signal at an adder 21 and the result is outputted as a television signal (y) subjected to noise reduction from a terminal 2. A 2-dimension spatial high pass filter is adopted as the high pass filter 51. Thus, the edge component and noise are separated.

Description

Detailed Description of the Invention [Technical Field of the Invention] The present invention relates to a television signal noise reduction device, and particularly to a television signal noise reduction device that reduces noise by using correlation between current and past frame image signals. Regarding equipment.

[Technical background of the invention]

A conventional noise reduction device (see @Digital Signal Processing of Images, published by Nikkan Kogyo, author: Takahiko Fukinuki, pH 5 to p. 118) is shown in FIG.

The current input television monitor signal X inputted from the terminal 1 is subtracted by the prediction signal from the prediction circuit 41 in the subtracter 11, and a prediction error signal ε(=仝−x) is obtained. This prediction error signal C is converted into a correction signal 61 by a correction signal generation circuit 31, and then converted into a correction signal 61 by an adder 21, which is converted into a correction signal 61 by an adder 21.
and is output from terminal 2 as a television signal y with reduced noise. Further, this television signal y is added to the prediction circuit 41, and a new predicted signal amount is obtained based on this.

Here, for example, the prediction circuit 41 for the NTSC signal is configured as shown in FIG. The noise-reduced television signal y is input from a terminal 411, delayed by a frame delay circuit 413, and guided to a Y/C separation circuit 414. A Y/C separation circuit 414 separates a luminance signal (Y) and a color signal (C). Color signal is chroma inverter 41
After the polarity is inverted at step 5, it is added to the luminance signal again at adder 416, and a predicted signal intensity is obtained from terminal 412.

In the circuit shown in FIG. 8, the input is X, the output of the prediction circuit 41 is y, the output of the adder 2 is y, and the correction signal ε° is
-x) multiplied by a coefficient K, then y can be found as follows.

Y =X +K (kai-X) ” (1-K)
) Mountain (1) Generally, the subject is often stationary in areas where x-xl is small, and is often moving in areas where x-xl is large. Therefore, for example, as shown in FIG. 10, when 1-x1 is small, choose K close to 1, and
．． When −xi is large, K=0, and in the stationary part
By outputting y (=X), which has less noise than y, and by outputting y (=X) in order to avoid afterimages in moving parts, it is possible to obtain y with improved image quality of input X.

Fig. 11 shows another conventional example, which is similar to Fig. 8 in that when obtaining a television signal y with reduced noise, a correction signal ε° is added to the predicted signal instead of the input television signal X. different from. Let the input be X, the output of the adder 2I be y, the output of the prediction circuit be 2, and the correction signal ε° be the product of (X−仝) multiplied by the coefficient (1−K). It is found that yKY=(1-K)(x-仝)+仝=(1-K)x+, and the same output as in FIG. 8 is obtained.

The transfer function of this noise reduction device is determined as follows from equation (13).

y = (1-K) x+xQ =(1-K)x+Kyz,' However, Z, -1 represents the frame delay.

” (1”Z7-')y=(1-K) x Therefore, the transfer function H(ZF-1) is expressed as. This characteristic is shown in FIG.

[Problems with background technology]

In the conventional apparatus, only the size of the difference C between the input television signal X and the predicted signal intensity is used to determine whether the difference is due to the motion K of the moving image or the noise of the still image. Therefore, C8, which defines the characteristics of the correction signal generation circuit shown in FIG.
By increasing the value of eo, it is possible to reduce noise, especially for videos that the human eye can follow, but from the characteristics shown in Fig. By narrowing the passband in the low-pass filter, even one image per K becomes blurred over time. Furthermore, when the value of eo is small, there is a drawback that noise cannot be sufficiently reduced.

[Purpose of the invention]

An object of the present invention is to improve the above-mentioned television signal noise reduction device, and to achieve a sufficient noise reduction effect without deteriorating the image quality, especially for moving images that the human eye can follow. The object of the present invention is to provide a noise reduction device that obtains the desired results.

[Summary of the invention]

In order to achieve the above object, the present invention focuses on the property that image noise occurs randomly not only temporally but also spatially, and uses a spatial filter to spatially separate image signal and noise components. It also distinguishes between images and reduces noise without deteriorating image quality.

[Embodiments of the invention]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Before describing one embodiment of the present invention, the present invention will be explained in detail below. Figure 6 shows the current input signal X when the object moves and the predicted signal from past images superimposed, and Figure 7 shows the signal amplitude between SS° in Figure 6. It is something. As is clear from FIG. 7, the prediction error signal x-A spreads out spatially with respect to the movement of the object. When viewed in the spatial frequency domain, this is mainly a low frequency component.

On the other hand, noise generation is random not only temporally but also spatially, which means that it mainly contains high-frequency components in the spatial frequency domain.

Focusing on this point, the present invention removes the part caused mainly by image movement from the prediction error signal using, for example, a two-dimensional spatial high-pass filter, and passes the part caused by noise, thereby improving the image quality. It reduces noise without deterioration.

FIG. 1 shows an embodiment of the present invention. The current input TV show signal X input from the terminal 1 is subtracted by the prediction signal from the prediction circuit 41 in the subtracter 11, and a prediction error signal ε is obtained. This prediction error signal C is passed through a high-pass filter 51.
, only the high frequency component εH is extracted and sent to the correction signal creation circuit 3.
1, it is converted into a correction signal gH1, which is added to the current input television signal in an adder 21, and outputted from a terminal 2 as a noise-reduced television signal y. This noise reduction signal y is also added to the prediction circuit 41 to obtain a new prediction signal.

Now, when performing noise reduction in the above configuration, it becomes a problem to distinguish between the edge portion of an image having high-frequency components in the spatial frequency domain and the original noise. Here, horizontal or vertical edge portions are particularly visually problematic, while diagonal edge portions do not pose a problem due to human visual characteristics.

In the spatial frequency domain, the horizontal edge is the shaded area A in the schematic spatial frequency characteristic diagram shown in Figure 2, and the vertical edge is the dotted area B in the same figure. . Therefore, the high-pass filter 51 used in the present invention is a two-dimensional spatial high-pass filter whose pass area is the perspective part C in FIG. 3, which does not pass through the areas A and H. As a result, the P-ji component and noise can be separated. Note that FIG. 4 shows an example of tap coefficients of a digital filter that realizes this characteristic. In the figure, the coefficients in the horizontal direction indicate the gain of each signal after being delayed by a unit pixel, and the coefficients in the vertical direction indicate the gain of each signal after being delayed by one horizontal line.

The implementation sequence of the present invention shown in FIG. 5 corresponds to the conventional example shown in FIG. 11, and the prediction error signal ε is guided to the high-pass filter 51 in the same manner as in the above-described embodiment.

〔Effect of the invention〕

As described above, the present invention distinguishes between noise and image signals not only temporally but also spatially, so that it is extremely effective even in moving images that the human eye can follow without deteriorating the image quality. can reduce noise.

[Brief explanation of drawings]

Fig. 1 is a block diagram of a noise reduction circuit according to an embodiment of the present invention, Fig. 2 is a schematic spatial frequency characteristic diagram showing the edge portion of a television broadcast signal, and Fig. 3 is a high frequency characteristic diagram included in the present invention. A schematic spatial frequency characteristic diagram showing the passband of a pass filter,
FIG. 4 is a numerical display diagram showing tap coefficients when configuring the high-pass filter, FIG. 5 is a configuration diagram showing another embodiment of the present invention, and FIGS. 6 and 7 are input television signals. 8 and 11 are diagrams showing the configuration of the conventional device, Figure 9 is the configuration diagram of the prediction circuit, Figure 10 is a characteristic diagram of the correction signal generation circuit, and Figure 12 is the noise diagram. FIG. 3 is a characteristic diagram of a reduced signal. DESCRIPTION OF SYMBOLS 11... Subtractor, 21... Adder, 31... Correction signal creation circuit, 41... Prediction circuit, 51... High pass filter, X... Input television signal, Kanayama Predicted signal, y...
・Noise reduction signal, ε...Prediction error signal, εH...
・High frequency component, εH'...correction signal. Agent Patent Attorney Nori Ken Yudo Hiroshi Uji 1E No. 31! 1 Volume 4 Figure 5 Figure 6 Figure 7 Figure 8 fcS9 Figure 10 Figure 12

Claims (1)

[Claims] a prediction circuit that predicts an incoming input television signal;
a subtracter that outputs a prediction error signal that is the difference between the prediction signal output from the prediction circuit and the input television signal;
A high-pass filter that extracts a spatial high-frequency component of the prediction error signal, and a correction signal generation circuit that generates a correction signal according to the high-frequency component, and the correction signal and the input television signal or A noise reduction device for a television signal, characterized in that either the correction signal or the predicted signal is guided to an adder, and a television signal with reduced noise is obtained from the output of the adder.