JPH04329073A - Noise removing circuit for television receiver - Google Patents

Abstract

PURPOSE:To hold the edge information of an image, to determine a first threshold in accordance with the noise level of the image and to adaptively remove the noise. CONSTITUTION:The number of picture elements in which the absolute value of each difference between the attention picture element in a window and other picture element is a first threshold S is obtained by an S deciding circuit 16, a K deciding circuit 20 decides whether or not the number of the picture elements is a second threshold K or below, and at the time of K or below, a mean value circuit 26 obtains the mean value of all picture elements in the window, when the number is not so, the mean value of the picture element of S or below and the attention picture element is obtained, and either of them is replaced with an original picture element value. An S setting circuit 18 determines S in accordance with the noise level for one of two field picture elements or above.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise removal circuit for a television receiver, and more particularly to a noise removal circuit for a television receiver that performs digital signal processing.

0002

2. Description of the Related Art In a television receiver, when noise is mixed into the receiver, from broadcast waves, from a VTR, etc., it appears as a random pattern of noise on the screen, which may deteriorate the image quality. If a pixel containing noise is included in a still area, a conventional noise removal method using pixels from adjacent frames can obtain an image without blur.

[0003] However, when pixels with noise mixed in are included in a moving image area, when processing is performed over frames as described above, correct pixel values cannot be obtained because the positions of the pixels are shifted. Therefore, it is necessary to perform processing using only pixels within the same frame. As a method for removing noise from such a moving image, there is a method using, for example, a simple average of about 5 to 9 surrounding pixels including the pixel of interest.

0004

However, this method has a problem in that the image becomes noticeably blurred because it removes even the high frequency components that appear at the edges of the original image. In order to reduce such blur, there is a method of increasing the weight of the pixel of interest when calculating the average value, but this has the opposite effect on noise reduction.

[0005] Therefore, the main object of the present invention is to provide a noise removal circuit for a television receiver that can effectively reduce noise and also reliably preserve edges.

[0006]

[Means for Solving the Problems] The present invention provides a television receiver in which a field image consisting of at least one signal component of component signals is
A window consisting of a pixel of interest and a plurality of pixels including the pixel of interest is set, and the absolute value of each difference between the pixel of interest in the window and other pixels in the window excluding the pixel of interest can be changed. Find the number of pixels that are below the threshold of
Then, if the number of pixels is less than or equal to a second variable threshold, the average value of all pixels in the window is calculated; otherwise, the average value is calculated only for pixels whose absolute values are less than or equal to the first threshold and the pixel of interest. , for a television receiver, which replaces one of the pixel values with the original pixel value, and includes a threshold value determining means for determining a first threshold value for each one or a plurality of field images.

[0007]

[Operation] Find the absolute value of each difference between the pixel of interest in the set window and other pixels in the window excluding the pixel of interest, and the absolute value is less than or equal to the changeable first threshold (S) Find the number of pixels (M). If the number of pixels (M) is less than or equal to a second variable threshold (K), calculate the average value of all pixels in the block; otherwise, the absolute value of the difference from the pixel of interest is less than or equal to the first threshold The average value of the pixel and only the pixel of interest is calculated, and one of them is replaced with the original pixel value to obtain a new pixel value. Therefore, when an edge exists within the window, M>K, and the average value of all pixels within the window is not replaced as a new pixel value, so the edge is reliably preserved.

[0008] Furthermore, for example, by detecting the noise level for each field image and determining the first threshold value (S) based on this, the optimum threshold value is used for each image. Noise is adaptively removed even when reception conditions differ.

[0009]

According to the present invention, since noise is removed while retaining the edge information of the image, a good television image can be obtained. Furthermore, since a threshold value determination means is used, noise can be removed adaptively, and S/
N ratio can be improved. The above objects, other objects, features and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the drawings.

0010

[Example] First, the principle of this invention will be explained. In one field of a raster-scanned television image, as shown in FIG. 2, the number of pixels in the horizontal direction of the display screen 1 is h, and the number of pixels in the vertical direction is v. NTSC
In the example of 4fSC sampling (fSC: color subcarrier frequency, approximately 3.58MHz) of the method, h=910, and v
=263.

FIGS. 3 and 4 show the correspondence between the virtual field screen 1 for processing and the pixel numbers of the odd and even fields that are actually interlaced scanned by the scan line 2. in the second half of the last line, and in even fields the first
There are no pixels in the front half of each line. However, for convenience of explanation, in the odd field, the leftmost pixel 2es (Fig. 3) of the first line is set as the starting point g00, and in the even field, the first line is virtually extended, and the virtual pixel 2os (Fig. 4) is set as the starting point g00. Similarly, the last line in the odd field is virtually extended to make pixel 2ee (FIG. 3) the end point gn.
−1 V−1, and in even fields, the rightmost pixel 2oe (Fig. 4) of the final line is set as the end point gn.
-1 Set as V-1.

In the field, as shown in FIG. 5, the number of pixels in the horizontal direction is (2n+1) and the number of pixels in the vertical direction is (2m+1).
+1) window 3 is set, the center pixel of window 3 is called the pixel of interest, and noise removal processing is performed on this pixel. In this example, n=m=2,
A window 3 consisting of 25 pixels (5 pixels x 5 pixels) is shown.

In this embodiment, a rectangular window as shown in FIG. 5 is set, but the window may have any shape, such as a polygon. Furthermore, the pixel of interest does not necessarily need to be at the center of the window. However, if the noise is spread randomly in two dimensions within the field, place the pixel of interest at the center of the window and set the window size to a rectangle.
It is effective to use a point-symmetrical shape such as a circle.

Let the pixel of interest be gij, and the pixel of interest gi
If the signal component at j is fij and the noise component is nij, then gij=fij+nij. However, i=0
,1,2,...,h-1, and j=0,1,2,...
..., v-1. The noise removal method according to the present invention calculates the absolute value of each difference between the pixel of interest and the pixel in the above-mentioned window 3, and calculates the absolute value of each difference between the pixel and the pixel of interest. The average value is calculated using only the average value.

If the pixel after noise removal is Fij shown in FIG. 6, Equation 1 can be obtained.

[0016]

[Math 1]

However, in Equation 1, (gij-S)≦
When gkl≦(gij+S), δkl=1, and (g
When ij−S)>gkl or (gij+S)<gkl, δkl=0. Although the value of the threshold value S depends on the image, it is appropriate to set the value to about 2 to 3% of the maximum possible value of the signal component. However, as will be explained later with reference to FIGS. 7 and 8, this threshold value S is determined for each field image based on the noise level.

Furthermore, in order to remove isolated points within window 3, if the number of pixels (denoted as M) that enter gkl±S is less than a separately set changeable threshold value (denoted as K), the pixel of interest g00 The noise in is determined to be spot noise, and the simple average of all pixels within the window is used as the new pixel value Fij. In an example where the window size is 5×5, it is appropriate that the threshold value K=3. In this case, M=0,1
, 2, 3, δkl is always 1, and the simple average of all pixels in the window is calculated, and when M = 4 to 24, δk
The average value is calculated using only the pixels for which l=1. If an edge component exists in the image, it will not become an isolated point within the window and M>K, so the edge can be preserved and an image with less blur will be obtained.

An actual circuit configuration based on such a principle will be explained with reference to the block diagram of FIG. Also in the embodiment shown in FIG. 1, for simplicity, the window size is set to a rectangle of 5 pixels x 5 pixels, and the pixel of interest g00 is set as the center pixel. However, as described above, the window size, shape, and position of the pixel of interest are arbitrary. In the embodiment shown in FIG. 1, a component signal such as a luminance signal is input to the input terminal 12, and the luminance signal is matrixed into a total of 25 pixels (5 pixels x 5 pixels) by a delay element such as a line memory in the window matrix 14. Ru. Image data from this window matrix 14 is input to an S determination circuit 16. This S determination circuit 16 is provided with the aforementioned threshold value S by an S setting circuit 18, which will be explained in detail later.
is set.

In this S determination circuit 16, the pixel of interest g
The difference between each of the 24 pixels excluding 00 and the pixel of interest is taken, and the absolute value of each difference is compared with the threshold value S set by the S setting circuit 18. Then, if the absolute value of each difference is within the threshold value S, the flag FSkl for that pixel is activated. Here, k=0, 1, 2, 3, 4, and l
=0, 1, 2, 3, 4. However, k=2 and l=
2 is excluded because it corresponds to the pixel of interest.

The K determination circuit 20 calculates the number M of pixels in which the flag FSkl is active among the 24 pixels. Here, M is compared with a changeable threshold value K set by the K setting circuit 22. Then, if M≦K, flag FK is activated. Note that the timing adjustment circuit 24 is for matching the processing time in the S determination circuit 16 and the K determination circuit 20 with the phase of the 5 pixel×5 pixel data from the window matrix 14, and mainly includes a memory.

Finally, the average value circuit 26 calculates the average value using the flags FK and FSkl. That is, when the flag FK is active, the average of all 25 pixels in the window is always calculated and set as a new pixel value Fij, regardless of the flag FSkl. However, when the flag FK is inactive, the average value is calculated only for pixels for which the flag FSkl is active and the pixel of interest, and the average value is calculated as a new pixel value Fij. This new pixel value Fij is output from the output terminal 28.

Noise removal is achieved by sequentially performing such processing on all pixels in the field, but the degree of noise may vary due to temporal changes in the reception status of broadcast waves and differences in reception status depending on the channel. Since the values differ, it is desirable to set the optimal threshold value S depending on the image. For example, in an image with a small S/N ratio, increasing the threshold value S slightly increases blurring, but it is better to strengthen the noise removal effect. On the other hand, for images with a large S/N ratio, a strong noise removal effect is not necessary, so the threshold S
By reducing the value, the deterioration due to image blurring can be reduced.

Therefore, the S setting circuit 18 shown in FIGS. 7 and 8 adaptively determines the threshold value S so as to obtain an image with less blur and a good S/N ratio even when the noise level changes over time. I have to. There are several methods for detecting the noise level in a field image, but in the embodiments shown in FIGS. 7 and 8, a method using the standard deviation of noise will be described. Since there is usually a strong correlation between pixels within a local area, it can be said that the standard deviation of the pixels within the 5×5 window described above approximately represents the standard deviation of the noise level within the window. However, this does not necessarily apply depending on the image, so specifically,
The in-window standard deviation is sequentially calculated for the entire field image, and the minimum value thereof is taken as the standard deviation of the noise level of the field image. However, the window size is not limited to 5×5, but if it is too wide, it is not preferable because the standard deviation of the image itself will be calculated or the amount of hardware will increase.

Referring to FIG. 7, an input component signal from a terminal 30 is converted into a 5×5 window matrix by delay elements such as a line memory 32 and a DFF (D flip-flop) 34, and an adder 36 converts the input component signal into a 5×5 window matrix. The sum of pixels (ADij) is calculated, and in order to obtain its average value, the divider 38 calculates (ADij÷25), which is taken as the average value (Xij). This average value is given to the difference square calculation circuit 44 shown in FIGS. 8 and 9.

Referring to FIG. 8, in order to obtain the sum of squared differences between each pixel within the window and the average value, first, the output of the timing adjustment circuit 40 (FIG. 7), that is, the output of the terminal 42, is The adder 46 and multiplier 4 of the difference square calculation circuit 44 are formed into a 5×5 window matrix using the line memory 32, DFF 34, etc., and each output is shown in detail in FIG.
8 to find the square of the difference (gkl-Xij)2,
An adder 50 calculates the sum of 25 pixels. This result,
After dividing by 25 by the divider 52, the square root calculation circuit 54
By finding the square root of the standard deviation (d
ij) is calculated. A series of calculations like this are performed for all pixels in one field, the minimum value is determined in the minimum value comparison circuit 56, and this is calculated as the standard deviation of the field image (
D). Since the standard deviation (dij) is obtained for each pixel clock, the minimum value can be calculated by successive comparisons for each field image.

Note that the threshold value S for the difference between the noise and the signal that is actually used is calculated by using this standard deviation (D) by the multiplier 58.
) multiplied by an integer (C) is used. However, the value of C may be about 2 to 3. The threshold value S thus obtained is input to the noise removal circuit 10 described above. However, in order to adjust the timing during calculation of the threshold value S, the component signal is applied to the input terminal 12 of the noise removal circuit 10 through timing adjustment circuits 40 and 60 made up of delay elements.

In the embodiments shown in FIGS. 7 and 8, dividers 38 and 52 are used, but these are used as a circuit that multiplies by (1/25) by a multiplier or as a table using a memory element such as a ROM. may be configured. Furthermore, when calculating the square root, a memory element such as a ROM may be used in addition to a dedicated arithmetic circuit such as the square root arithmetic circuit 54 described above.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention.

FIG. 2 is an illustrative diagram showing a display screen.

FIG. 3 is an illustrative diagram showing the correspondence between pixel positions in an odd field and pixel numbers in actual processing.

FIG. 4 is an illustrative diagram showing the correspondence between pixel positions in an even field and pixel numbers in actual processing.

FIG. 5 is an illustrative diagram showing a window and signal components of its pixel of interest.

FIG. 6 is an illustrative diagram showing a window and pixels after noise removal.

FIG. 7 is a block diagram showing part of the S setting circuit of the embodiment of FIG. 1;

FIG. 8 is a block diagram showing other parts of the S setting circuit of the embodiment of FIG. 1;

9 is a block diagram showing the difference square calculation circuit of FIG. 8; FIG.

[Explanation of symbols]

1...Field screen 2
...scan line 3
...window 10
...Noise removal circuit 12
...Input terminal 14 ...Window matrix 16 ...S judgment circuit 18 ...S setting circuit 20
...K judgment circuit 22
...K setting circuit 24, 40, 60 ...timing adjustment circuit 26 ...average value circuit 28 ...output terminal 30
...Input terminal 32
...Line memory 34...
DFF36, 46, 50... Adder 38, 52... Divider 42... Terminal 44... Difference square calculation circuit 48
, 58 ... multiplier

Claims (1)

[Claims] Claim 1: In a television receiver, a window consisting of one pixel of interest and a plurality of pixels including the pixel of interest is set in a field image consisting of at least one signal component of component signals, and Find the number of pixels for which the absolute value of each difference between the pixel of interest and other pixels in the window excluding the pixel of interest is less than or equal to a first changeable threshold, and a second threshold for which the number of pixels is changeable. If the absolute value is less than or equal to the first threshold, the average value of all pixels in the window is calculated, and if not, the average value is calculated only for the pixels whose absolute value is less than or equal to the first threshold value and the pixel of interest, and one of them is converted to the original pixel value. What is claimed is: 1. A noise removal circuit for a television receiver, comprising threshold value determining means for determining the first threshold value for each one or a plurality of field images.