JP2637978B2 - Motion adaptive image quality improvement device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a motion-adaptive image quality for suppressing noise components contained in an image signal and improving sharpness in a device such as a television receiver which handles an image. It relates to an improvement device.

(Prior Art) Conventionally, as an effective noise reducer for a moving image such as a television signal, spatio-temporal smoothing that changes the characteristics of noise reduction in a space domain and noise reduction in a time domain by adapting to a temporal change of an image. (See IEICE Transactions Vol. J70-B-1; Yashima, Sawada, "Intra-frame / inter-frame adaptive extrapolation and interpolation coding for HDTV signals").

The operation of this filter is to compare the signal of the previous frame processed by the time domain filter with the current frame signal, and if there is no difference, a spatial low pass filter (LP
By applying LPF up to a certain amplitude change when there is a difference without applying much to F), it is intended to reduce noise without causing motion blur even in a moving image.

(Problems to be Solved by the Invention) However, the above-mentioned conventional spatio-temporal filter has been mainly used as a pre-process in signal encoding to reduce noise. Therefore, contour compensation in a television receiver or the like is not considered. In the case of this processing, the sharpness is not increased even if the image is blurred.

In addition, since the adaptive processing is performed independently for each pixel based on the time change of each pixel, even in a still image, if there is much noise, it is detected as a motion particularly in pulse noise, and the noise is smoothed by a spatial filter. become.
Conversely, even in the moving area of the image, depending on the shape of the image, some pixels become still, and even when the contrast of the shape change is low, there is little change in the time of the pixel, and the image becomes still. I will.

From the above points, the conventional spatio-temporal filter tends to cause an inappropriate filter operation, and the resolution tends to decrease. In particular, when there is a lot of noise, the time-domain filter does not work effectively, which is a problem.

Therefore, an object of the present invention is to provide a motion-adaptive image quality improving device which solves the above-mentioned problems of the conventional technology.

(Means for Solving the Problems) In order to achieve the above object, the present invention provides a method for subtracting a current frame value from a current frame value through a subtraction circuit for subtracting the current frame value from the previous frame value processed in the time domain. A motion detecting unit that obtains a motion detection value through a first low-pass filter having a narrow band by converting a difference between each pixel to an absolute value, a first delay device that compensates for a delay in the first low-pass filter, A delay subtraction result obtained by delaying a subtraction result of the subtraction circuit through the first delay unit and the motion detection value are input, and an output gain corresponding to the input delay subtraction result is set to a value when the motion detection value is “0”. Sometimes, the maximum gain is given to form a low-pass filter with a narrow band, and as the motion detection value increases, the gain drops suddenly at a predetermined value or more and the limit-type filter is formed. A first nonlinear circuit for changing the gain in accordance with both input values and outputting the same, and a delay obtained by delaying the output of the first nonlinear circuit and the current frame value through the first delay unit. A time-domain filter for performing processing in the time domain by an adding circuit for adding an output to the current frame value, and a second low-pass filter in the spatial domain for processing output of the time-domain filter in the time domain; Second and third delay units for compensating for a delay, a subtraction circuit for subtracting a processing output in the time domain passed through the second delay unit from a value passed through the second low-pass filter, and the subtraction circuit And the motion detection value passed through the third delay unit are input, and the gain of the output with respect to the input subtraction result is set to a negative value when the motion detection value is "0" to make the contour compensation. Is done, A second non-linear circuit that increases the gain as the motion detection value increases so that the second low-pass filter operates, and changes and outputs the gain according to both input values; And a spatial-domain filter for performing processing in a spatial domain by an adder circuit for adding an output of the nonlinear circuit and a processing output in the time domain through the second delay unit to obtain an output. The present invention is to provide a motion-adaptive image quality improving apparatus that performs the following.

(Embodiment) The motion-adaptive image quality improvement apparatus according to the present invention appropriately detects motion in a moving image, performs sufficient noise reduction with a time-domain filter in a still region, and performs contour compensation with a spatial-domain filter. The sharpness of the image is increased, the noise reduction of the time domain filter is minimized in the motion area to prevent motion blur, and the noise is suppressed by applying a low-pass filter (LPF) with the spatial domain filter.

FIG. 1 is a block diagram showing an embodiment of a motion adaptive image quality improving apparatus according to the present invention.

In FIG. 1, an input terminal 1 receives an image signal such as a television signal. 2, 13 is a subtractor, 3 is a limiter,
4 is an absolute value (ABC) circuit, 5 and 11 are low-pass filters (LP
F).

D ₁ delay 6 is a delay amount D ₁ for compensating for the delay of LPF _5.
It is a delay device. Reference numerals 7 and 14 denote nonlinear circuits, and reference numerals 8 and 15 denote adders.

(1F-D ₁₎ Delay 9 is the only delay D ₁ from one frame less delay amount of the delay device.

D ₂ delay 10 and 12 is a delay unit of the delay amount D ₂ to compensate for the delay of the LPF 11. 16 is an output terminal.

The first half (left half of the figure) of the device constitutes a time-domain filter, and is basically a recursive filter having a one-frame delay. The second half (right half of the figure) of the device constitutes a spatial filter and has a one-dimensional or two-dimensional LPF 11.
Also F. Also, D ₁ delay 6 and D ₂ delay 10, 12
Compensates for the delays of LPF5 and LPF11 respectively.
This is for maintaining the phase relationship of the LPF correctly.

Next, the operation of each unit in the above configuration will be described.

First, as for the time-domain filter, the value of the current frame is subtracted for each pixel by the subtractor 2 from the value of the previous frame that has been cyclically added, to obtain a difference. This difference component is passed through the D ₁ delay 6 to compensate for the delay of LPF5 the motion detection circuit described later. Further, same applies to the current frame values for matching the timing with the addition of post-transparent to D ₁ delay 6.

Difference signal, up frame through D ₁ delay 6, the following process is the non-linear circuit 7. First, in the stationary region, a state close to a gain of “1” with a linear output is used to enhance the effect of the noise reducer. As a result, an LPF having a considerably narrow band is formed in the time-domain filter, and random noise is greatly reduced. On the other hand, in the motion area, a limit type characteristic which is generally used is used, and the gain is suddenly decreased at a certain value or more as the input value increases, so that blurring of the motion is prevented. FIG. 2 shows an example of the input / output characteristics of the nonlinear circuit 7. here,
The load is symmetric. The nonlinear circuit 7 has a limiter internally. When the input signal level is about "0 to 255", a limiter is applied at about "64", and the same processing is performed for a higher level. Here, usually, when the difference between frames is large, it is a motion area. In that case, the characteristic of the nonlinear circuit is that all outputs are “0” at a certain input value or more. There is no.

The output of the nonlinear circuit 7 obtained in this way and the current frame value are added by the adder 8 to obtain an output with reduced noise.

Next, regarding the spatial filter, the LPF 11 changes the two-dimensional characteristics of the filter, such as operating two types of filters, vertical and horizontal, separately according to the shape of the image in space. Can be used, but here, there is no two-dimensional characteristic change, and the characteristic change is one-dimensional.

The configuration of the spatial bandpass filter is based on LPF1
From multiplied by 1, D ₂ to compensate for the delay of the LPF11
The input signal value passed through the delay 12 is subtracted by the subtractor 13.
Here, the LPF 11 may have a one-dimensional configuration or a two-dimensional configuration, and in the case of a two-dimensional configuration, more ideal characteristics can be obtained.
An example of a simple one is shown in FIG. In FIG. 5, 1H delay is a delay unit for one line, T is a delay unit for one dot, and an adder.

FIG. 7 shows the pixel coefficients of the secondary LPF in FIG. 5 corresponding to the secondary pixel arrangement, and the pixel coefficients (“1”, “2”, or “4”) shown in FIG. Multiplied by 1/16 each.

As described above, since the value to which the LPF 11 is not applied is subtracted from the output of the LPF 11 by the subtractor 13, a difference is obtained as a frequency range, and a high frequency component of the signal is obtained. This high frequency component is passed through the non-linear circuit 14. If the non-linear circuit 14 is linear and has a gain of "1", the value of only the delay is added by adding the value of only the delay later. The output of the LPF 11 is cancelled, and the output of the LPF 11 is output as it is. On the other hand, if the output of the nonlinear circuit 14 is "0", the value of only the delay is output as it is, and no processing is performed. Conversely, if the gain of the nonlinear circuit 14 is negative, the high-frequency components of the signal will be added in a positive phase, which is opposite to the case where the above-described gain is positive and the high-frequency components are reversed. , The high-frequency component of the output signal increases, and contour compensation is performed.

Therefore, the input and output polarities of the nonlinear circuit 14 may be made positive when the LPF 11 is to be applied, and may be negative when the contour compensation is to be performed. When the nonlinear circuit 14 is used, such a change in the processing can be set according to the level of the high frequency component,
This determines the non-linear characteristics.

Here, depending on how this processing is performed, LPF11 is used to reduce noise in places where the image is spatially flat and there is little change in the high frequency range. Where there are many area components, contour compensation is performed to increase the sharpness of the image. However, an image becomes unnatural if contour compensation is performed too much. Therefore, a limiter is applied to a value that changes the image by contour compensation.

Next, adaptive processing in a still area and a moving area is described.
Since the noise component is sufficiently suppressed by the time-domain filter in the stationary region, the LPF 11 is used only when the high-frequency component is relatively small, and in other cases, the contour compensation is applied relatively strongly.

Since the noise is sufficiently small, an image with high sharpness can be obtained without noticeable noise by the contour compensation. Conversely, in the motion area, noise reduction by the time-domain filter is not sufficient, and considerable noise is included, so LPF11 is applied to the high-frequency component level with relatively large amplitude, and contour compensation is not performed much. . In this case, the image is slightly blurred and the sharpness is lowered, but since the addition is moving, the high-frequency component is not so large in the first place, and it is difficult to detect visually, so this is not a major problem.

By performing the adaptation processing based on such a movement in multiple stages and making a smooth change, the image quality can be improved without any unnaturalness. FIG. 3 shows an example of the input / output characteristics of the nonlinear circuit 14 in the above-mentioned space region. Here, the negative direction of the input is symmetrical, the input absolute value is "32", and a limiter is applied. This is done so as not to make a significant change to the signal and to facilitate processing of the non-linear characteristics.

Note that these non-linear characteristics are stored in ROM (Read Only Memory).
Even if it is attempted to perform a table lookup (Table Look Up), when the motion adaptation is set to 4 levels, the capacity of the ROM can be as small as 4K bits in the nonlinear circuit 7 and about 1k bits in the nonlinear circuit 14. Become.

Next, detection of a moving area or a still area as a reference for adaptive processing in the time domain and the spatial domain as described above will be described.

The device of the present invention is fundamentally different from the conventional device in that a relatively strong correlation is provided for each pixel by applying a narrow-band LPF 5 to the motion detection output.
Normally, the motion in an image does not greatly change from pixel to pixel and has a fairly strong correlation. Therefore, a strong correlation is also given to the motion output, and the erroneous detection due to the noise component and the erroneous detection when the image having the high-frequency component in the spatial region moves are greatly improved compared to the case where the erroneous detection is performed for each pixel. Things.

The operation is to first detect a difference signal between frames. Since this signal is obtained in the time-domain filter as the output of the subtracter 2, it can be used as it is. Therefore, a simple limiter is applied to the output of the subtracter 2 by the limiter 3 and further converted to an absolute value by the absolute value (ABS) circuit 4, thereby obtaining a motion for each pixel. The output of the ABS circuit 4 is passed through an LPF 5 having a considerably narrower band to give a strong correlation, and even if the motion of each pixel fluctuates, the information of the final operation area or the still area is not very much for each pixel. Make no change.

Here, the limiter 3 is used to reduce erroneous detection of a static region as a moving region due to strong pulsed noise, and to reduce a difference in motion detection level due to a difference in the degree of change in signal (contrast ratio). used. Here, although the latter reason is a similar motion area, linear motion detection is detected in proportion to the contrast ratio of the spatial change in the image, but the ratio is controlled to some extent. This is to make it smaller than the ratio.

On the other hand, by inserting such a limiter 3, ABS
This is also useful for reducing the number of processing bits in the digital processing of the LPF 5 and the absolute value by the circuit 4. FIG. 4 shows an example of the characteristics of the limiter 3.

What matters here is the characteristics of the LPF 5, but considering the characteristics of the image, it is desirable that the LPF 11 has a narrower band than the LPF 11 used in the spatial domain.

FIG. 6 is a diagram showing the configuration of LPF5.
The 1H delay is a delay for one line, T is a delay for one dot, 9T is a delay for nine dots, and an adder.

In the example of the LPF 5 shown in FIG. 6, the vertical direction is a spatial area.
Same as LPF11, but with the same coefficient form in the horizontal direction, the bandwidth is narrowed by a simple circuit by adding, delaying and then subtracting.

FIG. 8 shows the pixel coefficients of the secondary LPF in FIG. 6 corresponding to the secondary pixel arrangement, and the pixel coefficients ("1") shown in the figure are each multiplied by 1/27.

Since LPF5 here does not directly change the image quality unlike LPF11 in the spatial area, it is not very strict, and what is optimal considering the various movements of the image It is difficult to judge.

(Effects of the Invention) As described above, with the above-described motion adaptive image quality improvement device of the present invention, when the noise component is relatively large in the related art, the noise is not so much reduced by the time-domain filter even in the still image. Noise is sufficiently reduced by determining that a certain area is a stationary area. In addition, noise with respect to pulse noise is similarly reduced by the time-domain filter. In such cases, the spatial filter is not very
Since LPF does not act, the sharpness does not decrease much. Further, in the apparatus according to the present invention, when it is determined that the area is a still area, contour compensation works, and not only noise is reduced, but also the sharpness of an image can be improved. In this case, since the noise-reduced signal is processed, the noise is not conspicuous due to the contour compensation, and an overall excellent image quality can be obtained. On the other hand, when it is determined that the moving area is the moving area, the time domain filter performs the same operation as in the related art, and there is no significant problem even if the still area is erroneously determined to be the moving area. Furthermore, in a motion area, noise reduction is mainly performed in a spatial area, and finally, much noise does not remain in an image.

In addition, each process changes smoothly continuously,
No unnaturalness due to switching occurs.

These adaptive processes are less likely to malfunction even with pulsed noise and the like, and the erroneous determination of a moving area and a still area due to the spatial shape and contrast ratio of an image is reduced. Further, since this processing continuously changes for each pixel, the processing is greatly different between adjacent pixels, and the unnaturalness due to the processing changing at the boundary between blocks, such as detection in block units, is also eliminated.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a motion-adaptive image quality improving apparatus according to the present invention, and FIGS. 2 and 3 are diagrams showing input / output characteristics of a nonlinear circuit constituting an embodiment of the apparatus of the present invention. ,
FIG. 4 is a view showing characteristics of a limiter constituting one embodiment of the apparatus of the present invention, FIGS. 5 and 6 are views showing a construction of an LPF constituting one embodiment of the apparatus of the present invention, FIGS. FIG. 8 is a diagram showing the pixel coefficients of the secondary LPF in FIGS. 5 and 6, respectively. 1 ...... input terminal, 2,13 ...... subtractor, 3 ...... limiter, 4 ...... absolute value (ABS) circuit, 5,11 ...... low pass filter (LPF), 6 ...... D ₁ delay, 7,14 ...... nonlinear circuits, 8, 15 ...... _{adder, 9 ...... (1F-D 1} ) delay, 10, 12 ...... D ₂ delay, 16 ...... output terminal.

Claims (1)

(57) [Claims] An absolute value of a pixel-by-pixel difference between a previous frame value and a current frame value through a subtraction circuit for subtracting the current frame value from a previous frame value processed in the time domain is obtained. Motion detection means for obtaining a motion detection value through a low-pass filter; a first delay device for compensating for a delay in the first low-pass filter; and a subtraction result of the subtraction circuit delayed by passing through the first delay device. The delay subtraction result and the motion detection value are respectively input, and the output gain for the input delay subtraction result is given the largest gain when the motion detection value is "0", thereby forming a low-pass filter with a narrow band. So that as the motion detection value increases, the gain suddenly decreases above a predetermined value so as to have a limit type characteristic, and the gain is changed according to both input values. A first non-linear circuit that outputs the current frame value and a delayed current frame value obtained by delaying the current frame value through the first delay unit to obtain an output. A second and a third delay unit for compensating a delay of a second low-pass filter in a spatial domain with respect to a processing output in a time domain of the time-domain filter; A subtraction circuit for subtracting the processing output in the time domain passed through the second delay unit from the value passed through the low-pass filter, and a subtraction result of the subtraction circuit and a motion detector passed through the third delay unit. Are input, and the gain of the output with respect to the input subtraction result is made negative when the motion detection value is “0” so that contour compensation is performed. As the motion detection value increases, the gain is increased. big The second low-pass filter is actuated to change the gain in accordance with both input values, and outputs a second nonlinear circuit; and outputs the output of the second nonlinear circuit and the second delay unit. A motion-adaptive image quality improving apparatus, comprising: a spatial domain filter for performing processing in a spatial domain by an addition circuit for obtaining an output by adding a processed output in a passed time domain.