JP2573719B2 - Noise reduction device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention uses a field correlation or a frame correlation of a video signal of a television, a video, a video camera, or the like, thereby deteriorating an afterimage characteristic even in a moving image. Related to a device that reduces noise without
More specifically, the present invention relates to a noise reduction device capable of performing the best noise reduction for noise appearing at an edge portion of an input image during motion by providing an optimum value of a nonlinear processing method used in this system.

2. Description of the Related Art As a conventional noise reduction device, for example,
The thing as shown in 158574 is known.

FIG. 5 is a block diagram showing a conventional noise reduction device. In FIG. 5, 1 is an input terminal for a video signal, 2 is an A / D converter for converting an input video signal from an analog signal to a digital signal, and 3 is noise from an input video signal of one input signal to a noise of the other input signal. A subtraction circuit 4 for subtracting a signal is a frame memory for delaying the output signal of the subtraction circuit 3 in which noise is reduced from the input video signal by one frame. NTSC
The color signal of the color video signal is inverted for each frame, but reference numeral 5 denotes a color signal phase shift circuit for compensating the phase, and only the color signal of the video signal delayed by the frame memory 4 is inverted. I do. Reference numeral 6 denotes a subtraction circuit for subtracting an input video signal and a video signal delayed by one frame to obtain a difference frame difference signal, and 7 converts a frame difference signal output from the subtraction circuit 6 from a serial digital signal to a parallel digital signal. A serial-parallel converter 8 is a serial-parallel converter 7.
This is a Hadamard transform circuit that performs a Hadamard transform, which is a direct transform, on a parallel digital signal output from, and extracts a vertical component, a horizontal component, and an oblique component from the frame difference signal. The output of the Hadamard transform circuit 8 has a different distribution of the noise component and the motion component, and reference numeral 9 denotes a non-linear processing circuit that extracts the noise component by using the difference. Numeral 10 denotes an inverse Hadamard transform circuit. Since the noise signal extracted from the non-linear processing circuit 9 is obtained by Hadamard transform, the noise signal is returned to the original time axis by Hadamard inverse transform. Reference numeral 11 denotes a parallel-serial conversion circuit for converting a parallel digital signal into an original serial digital signal, and reference numeral 12 denotes a D / A converter for converting a digital signal output from the subtraction circuit 3 into an analog signal.

The operation of the thus configured conventional noise reduction device will be described. When a video signal is input from an input terminal 1, it is converted into a digital signal by an A / D converter 2, and the digital signal passes through a subtraction circuit 3 to subtract a non-correlation component described later, and ideally contains a noise component. No video signal component is stored in the frame memory 4 and delayed for one frame. Since the phase of the color signal of the video signal delayed by one frame is inverted with respect to that of the one frame signal, the phase is compensated by the color signal phase shift circuit 5 so that only the phase of the color signal is inverted. 6, a frame difference signal of the difference between the two video signals having the same chroma phase is obtained. Originally, when the input video signal is a still image, this frame difference signal becomes a noise component itself, and noise can be extracted without the need for a circuit described below. However, if the input video signal is a moving image, the frame difference signal is a signal in which a signal component having no frame correlation and a noise component are combined.

Next, a method of obtaining only a noise component from the frame difference signal will be described. This frame difference signal is converted from a serial digital signal to a parallel digital signal by a serial-parallel conversion circuit 7, and a Hadamard conversion circuit 8
The signal is divided into components such as a vertical direction component and a horizontal direction component that well represent characteristics as signals. Now, it is assumed that the transformation order of the Hadamard transform is a 4 × 2 order two-dimensional Hadamard transform. In this case, one pattern picture element of the input is as shown in FIG. 6, and the input (the next input area need not be limited to that shown in the figure). Becomes The converted output of the 2 × 4 order Hadamard transformation to F _24.

F ₂₄ = H ₂ · X ₂₄ · H _{4 From the} above equation, the output F _{24 of the} Hadamard transform is obtained from the 4 × 2 order input picture element X ₂₄ . H ₂ and H ₄ are as follows.

The output from the Hadamard transform circuit 8 is a converted output of eight components. On the other hand, since the noise has no correlation, the noise is almost evenly distributed among the eight component frequencies of the output of the Hadamard transform circuit 8. Since the noise level at the output of the Hadamard transform circuit 8 corresponds to the noise level of the input signal as is well known, only a low-level noise component can be extracted from each of these components through the nonlinear processing circuit 9. . Each component extracted by the non-linear processing circuit 9 is obtained by Hadamard transform,
By passing through the Hadamard inverse transform circuit 10, the signal is returned to the original time axis, and a parallel digital noise signal is obtained. Then, the parallel-serial conversion circuit 11 converts the parallel digital noise signal into a serial digital signal similar to the input form. The signal obtained here is a signal obtained by extracting only the noise component from the frame difference signal having no frame correlation, and is supplied to the subtraction circuit 3 as described above, and the noise-free digital signal is obtained by subtracting the noise component from the input video signal. A video signal is obtained. Finally, the A / D converter 12 converts the digital video signal into an original analog signal and outputs it.

In principle, the noise reduction apparatus using such a method has a possibility that noise can be reduced not only in a still image in inputting a video signal but also in a moving image without deteriorating the input video so much.

Problems to be Solved by the Invention The noise reduction device described in the related art can reduce noise without generating an afterimage not only in a video signal input and a still image but also in a moving image in principle.

However, the S / N improvement effect and the afterimage characteristic of the edge part visually differ greatly depending on the processing method of the nonlinear processing, and the optimal characteristic of such a nonlinear characteristic has not been proposed so far.

The present invention improves such a point, and provides a noise reduction device having a small degradation of an afterimage characteristic and a large S / N improvement effect by providing an optimal characteristic of the nonlinear processing method of the noise reduction device described in the conventional example. The purpose is to provide.

Means for Solving the Problems To solve the above problems, a noise reduction device of the present invention takes a difference component between an input video signal and an output signal of an n (n> 0: n is an integer) field delay unit. First subtraction means, feature extraction means for decomposing the output of the first subtraction means into a plurality of (N (N （1)) feature components related in the vertical and horizontal directions, and a plurality of features of the feature extraction means A plurality of non-linear processing means for extracting a noise component from each component output; a second subtraction means for obtaining a difference signal between an output from the non-linear processing means and an input video signal; and an output of the second subtraction means for n fields. The n-field delay means for delaying the signals by an amount equal to or less than a reference level an (an> 0) and a reference level bn (bn <0). Is proportional to the input signal Outputs an output signal proportional to a number in (however, in> 0), outputs a substantially constant value An when the input signal is higher than the reference level an, and outputs a substantially constant value Bn when the input signal is lower than the reference level bn. And the reference levels an and bn for the feature component representing the vertical motion component among the plurality of nonlinear processing means are relatively larger than the values for the other feature components, and the horizontal motion component is Reference level an, bn for the feature component to be represented
Is set to be relatively smaller than the values for the other feature components.

According to the above configuration, the nonlinear processing method for performing nonlinear processing on the output of the feature extraction circuit is performed when the input signal has the reference level a (a> a).
0) and outputs an output signal proportional with proportionality constant i is less and the reference level b (b> 0) or higher (where, i> 0), the reference level a higher outputs a substantially constant value A _n, or reference When the level is equal to or lower than the level b, the processing is performed by a non-processing circuit having an optimal non-linear characteristic that outputs a substantially constant _Bn . Therefore, it is possible to reduce noise and reduce noise with a large S / N improvement effect.

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

FIG. 1 is a block diagram of a noise reduction device according to a first embodiment of the present invention. In FIG. 1, components denoted by the same reference numerals as those in FIG. 5 are the same as those in FIG. 5, and have the same functions. Reference numeral 21 denotes a non-linear processing circuit which performs non-linear processing on the output from the Hadamard transform in the form of optimal characteristics, and is interposed between the Hadamard transform circuit 8 and the Hadamard inverse transform circuit 10 instead of the conventional non-linear processing circuit 9. Has been used.

The operation of the noise reduction device of the first embodiment configured as described above will be described below. When a video signal is input from the input terminal 1, the A / D converter 2 outputs 4 fsc (3 fsc
Is also considered), and is converted from an analog signal to a digital signal. The non-correlation component is subtracted by subtracting the video signal one frame before from the input video signal by the subtraction circuit 3, and the video signal ideally contains no noise component. This video signal component is stored in the frame memory 4 and is delayed for one frame. NT
Since the color signal of the SC color video signal is inverted for each frame, the color signal is compensated by the color signal phase shift circuit 5 and only the color signal of the video signal delayed by the frame memory 4 is inverted. The subtraction circuit 6 obtains a frame difference signal representing the difference between the two video signals. This frame difference signal is a signal in which a signal component having no frame correlation (that is, a motion component) and a noise component are combined. The frame difference signal is converted from a serial digital signal to a parallel digital signal by a serial-parallel conversion circuit 7, and a Hadamard conversion circuit 8, which is a signal feature extracting means, converts the frame difference signal into a low frequency component, a vertical frequency component, It is divided into signal components representing features such as horizontal frequency components. Since each of these components is a noise component and a motion component, only a low-level noise component can be extracted from these components by passing through the non-linear processing circuit 21. Then, the nonlinear processing circuit 21
Since each component extracted is obtained by the Hadamard transform, it is returned to the original time axis by passing through the inverse Hadamard transform circuit 10, and a parallel digital noise signal is obtained. Next, in the parallel-serial conversion circuit 11, the parallel digital noise signal is converted into a serial digital signal similar to the input form. The signal obtained here is a signal obtained by extracting only the noise component from the frame difference signal. Then, as described above, this signal is supplied to the subtraction circuit 3 and subtracted from the input signal, whereby a digital signal without noise can be obtained. Finally, the A / D converter 12 converts the digital video signal into an original analog signal and outputs it.

This embodiment is also possible using a field memory. Also, it is not necessary to limit the present invention to the NTSC composite signal, and it is apparent that the signal can be constituted by a baseband signal. In this case, the color signal phase shift circuit becomes unnecessary.

The non-linear processing circuit 21 of the noise reduction device having the above configuration will be described in detail. Generally, there are various characteristics of the nonlinear processing circuit as shown in FIGS. 2 (a) to 2 (d). A feature of the nonlinear processing that can be considered in principle is that, in principle, a positive output is performed for a positive input, and a negative output is performed for a negative input. For these nonlinear processes, numerical S / N improvement can be expected in principle. However, experiments by the present inventors have revealed that there is a relationship between the nonlinear characteristics and the visual evaluation of the afterimage characteristics. Here, the results will be described below.

The processing shown in FIG. 2 (a) is to output an output level proportional to an input component by a proportional constant k (k> 0). According to an experiment using this nonlinear processing circuit, an afterimage phenomenon occurs at an edge portion having a small level difference, and there is a problem that the edge is blurred and not clear.

The non-linear processing shown in FIG. 2 (b) is a substantially sine curve, and outputs a 0 level at a certain level or higher or below a certain level. According to an experiment using this non-linear processing circuit, there is a problem that an edge having a small level difference is easily blurred, and that noise flickers near the edge.

The non-linear processing shown in FIG. 2 (c) outputs an output level proportional to the input level with a proportional constant n (n> 0) at a certain input level, and outputs a level higher than a certain level or below a certain level. Proportional constant (m <
An output level proportional to 0) is output, and above a certain level or below a certain level, a 0 level is output. According to an experiment using this nonlinear processing circuit, FIG. 2 (b)
Almost the same as, edges with small level differences are easy to blur,
There is also a problem that noise flicker is observed near the edge.

The non-linear processing shown in FIG. 2 (d) outputs an output level proportional to the input level with a proportionality constant i (i> 0) up to a certain input level, and a substantially constant value A above the reference level a. Or c (i> | c |> 0) for the input signal
And outputs a value that is substantially constant below the reference level b or a value proportional to the input signal by d (i> | d |> 0). According to an experiment using this nonlinear processing circuit, it was confirmed that there was no blurring of the edge due to the afterimage phenomenon observed in the processing circuit described above, and that there was no flickering of noise at the edge portion. In this way, for those which are considered to be effective in principle, it has been confirmed from experiments by the present inventor that the nonlinear processing method of FIG. 2 (d) gives visually the best results.

Therefore, the present invention proposes the characteristic (d) as an optimal characteristic as a nonlinear processing circuit.

FIGS. 3 and 4 show a circuit diagram of a noise reduction device according to a second embodiment of the present invention and a detailed diagram of a main part thereof. 3 and 4, the same reference numerals as in FIG. 5 denote the same parts as in FIG. 5, and perform the same functions.

In FIG. 3, reference numeral 31 denotes a Hadamard conversion circuit group 31 provided between the serial / parallel switching circuit 7 and the parallel / serial conversion circuit 11, and as shown in FIG. 4, the Hadamard conversion circuit 8 and the Hadamard conversion circuit The nonlinear processing circuit 32 includes a nonlinear processing circuit group 32 to which a plurality of outputs are input from the circuit 8 and a Hadamard inverse conversion circuit 10 to which respective outputs of the nonlinear processing circuit group 32 are input. , And performs different nonlinear processing on each of the plurality of outputs from.

Each nonlinear processing circuit of this nonlinear processing circuit group 32 has:
As described in the first embodiment, the processing method of FIG. 2D is the most effective. The relative input level each of the non-linear processing circuit which is respectively output an output value proportional with proportionality constant _{i n (i n> 0)} , to a substantially constant value A _n or the input signal at the reference level a _n or And outputs a value proportional to c _n (i _n || c _n |> 0). At a reference level b _n or less, a substantially constant value B _n or d _n (i _n > | d _n |> This is a non-linear processing circuit that outputs a value proportional to 0).

The operation of the thus configured noise reducing device of the present embodiment will be described below. When a video signal is input from the input terminal 1, it is converted from an analog signal to a digital signal by the A / D converter 2. The non-correlation component is subtracted by subtracting the video signal one frame before from the input video signal by the subtraction circuit 3, and the video signal ideally contains no noise component. This video signal component is stored in the frame memory 4 and is delayed for one frame. Since the color signal of the NTSC color video signal is inverted for each frame, the color signal 5 is compensated by the color signal phase shift circuit 5 and only the color signal of the video signal delayed by the frame memory 4 is inverted. The subtraction circuit 6 obtains a frame difference signal representing the difference between the two video signals. This frame difference signal is a signal in which a signal component having no frame correlation (that is, a motion component) and a noise component are combined. The frame difference signal is converted from a serial digital signal to a parallel digital signal by a serial-parallel conversion circuit 7, and a Hadamard conversion circuit 8 converts the frame difference signal into a reduced frequency component, a vertical frequency component,
It is divided into a plurality of components such as horizontal frequency components. Since these components are a noise component and a motion component, by passing each component through the non-linear processing circuit group 32, only a small-level noise component can be extracted from each component.

Here, the characteristics of the Hadamard transform output will be described in detail.
Generally, a video signal has a feature that the movement in the horizontal direction is large and the movement in the vertical direction is small. Therefore, for the vertical motion component, it is preferable to extract most of the noise component to the nonlinear circuit output even if the motion component is slightly mixed. Therefore, it is desirable to set the reference levels b _n and c _n high, and the afterimage degradation in the vertical direction is slightly increased, but there is no problem in the output image, and high S / N improvement can be performed. Conversely, it is better for the non-linear processing circuit to reduce the noise component extraction capability for the horizontal motion component even if the noise component extraction capability is reduced. Therefore, it is desirable to set the reference levels b _n and c _n low. As a result, the S / N improvement ability is reduced, but the afterimage phenomenon can be suppressed.

As described above, from each of the components decomposed into the Hadamard-transformed feature components, non-linear processing of different characteristics according to each feature component is performed, and a noise component is extracted. And
Since each component extracted from each of the nonlinear processing circuits is obtained by the Hadamard transform, it is returned to the original time axis by passing through the inverse Hadamard transform circuit 10, and a parallel digital noise signal is obtained. Next, in the parallel-serial conversion circuit 11, the parallel digital noise signal is converted into a serial digital signal similar to the input form. The signal obtained here is a signal obtained by extracting only the noise component from the frame difference signal. Then, as described above, this signal is supplied to the subtraction circuit 3 and subtracted from the input signal to obtain a digital signal free of noise. Finally, the A / D converter 12 converts the digital video signal into an original analog signal and outputs it.

This embodiment is also possible using a field memory. Also, it is apparent that the present invention is not limited to the NTSC composite signal, but can be configured with a baseband signal. In this case, the color signal phase shift circuit becomes unnecessary.

Further, although the Hadamard transform is used in the embodiment, a discrete cosine transform (DCT) may be used. In this case, the inverse Hadamard transform in the embodiment is an inverse discrete cosine transform (inverse DCT).

Advantageous Effects of the Invention As described above, according to the present invention, the limiter characteristics are provided as the optimal characteristics of the nonlinear processing performed on the signal components whose characteristics have been extracted, and the limiter characteristics are made different for each characteristic component, thereby making it relatively simple. With such a configuration, it is possible to obtain a noise reduction device with little degradation of afterimage characteristics and a large S / N improvement effect.

[Brief description of the drawings]

FIG. 1 is a block diagram of a noise eliminator according to a first embodiment of the present invention, FIGS. 2 (a) to 2 (d) are characteristic diagrams of a nonlinear circuit, and FIG. 3 is a second embodiment of the present invention. FIG. 4 is a block diagram showing details of a Hadamard transform circuit group in the noise removing device, FIG. 5 is a block diagram of a conventional noise reducing device, and FIG. 6 is a pixel of an input signal. It is a figure showing a unit. 1 ... signal input terminal, 2 ... A / D converter, 3, 6 ... subtraction circuit, 4 ... frame memory, 5 ... color signal phase shift circuit, 7 ... series-parallel conversion circuit, 8 ... ... Hadamard conversion circuit, 10 ... Hadamard inverse conversion circuit, 11 ... Parallel-serial conversion circuit, 12 ... D / A converter, 13 ... Output terminal, 21 ...
Non-linear processing circuit, 31 ... Hadamard transformation processing circuit group, 32
... A group of non-linear processing circuits.

Claims (1)

(57) [Claims] 1. An input video signal and n (n> 0: n is an integer)
First subtraction means for taking a difference component from the output signal of the field delay means; and a feature for decomposing the output of the first subtraction means into a plurality (N (N ≠ 1)) of characteristic components related in the vertical and horizontal directions. Extracting means; a plurality of nonlinear processing means for extracting a noise component from each of a plurality of feature component outputs of the feature extracting means; a second subtracting means for obtaining a difference signal between an output from the nonlinear processing means and an input video signal; The n-field delay means for delaying the output of the second subtraction means by n fields, and the plurality of non-linear processing means are arranged such that an input signal of each non-linear processing means is equal to or lower than a reference level an (an> 0). And reference level
When bn (bn <0) or more, the proportional constant in
Outputs a proportional output signal (where: in> 0), outputs a substantially constant value An when the input signal is higher than the reference level an, and outputs a substantially constant value Bn when the input signal is lower than the reference level bn. And wherein the reference levels an and bn for the feature component representing the vertical motion component among the plurality of nonlinear processing means are relatively larger than the values for the other feature components, and represent the horizontal motion component. A noise reduction device, wherein the values of the reference levels an and bn for components are set relatively smaller than the values for other feature components.