JPS6297075A - Eliminating device for vertical stripe noise - Google Patents

Abstract

PURPOSE:To eliminate vertical stripe noises due to sampling pulses by extracting the picture element producing the vertical stripe noise by a convolutional operation and correcting only the signal level corresponding to the extracted picture element by another convolutional operation. CONSTITUTION:A coefficient equation IV is obtained by subtracting each coefficient of an equation II from that of an equation I. The equation IV is defined as Wkl and an arithmetic equation is executed for fij. Thus Gij = x of an equation III is obtained just with a single convolutional operation. Thus the coefficient of the equation IV is also set previously to a data unit. Then a binary picture is produced by the coefficient of the equation II and then a single convolutional operation is carried out by means of the coefficient of the equation IV. Thus Gij of the equation III is obtained. Therefore it is possible to obtain a picture free from vertical stripe noises by rewriting the fij on a picture 1 with the Gij obtained from the equation III if the corresponding j-th value of the i-th raster on the binary picture is equal to 1.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a device for removing vertical stripe noise that occurs when a video signal is digitized.

[Scenery of invention]

As shown in FIG. 2, one raster of the video signal generated by scanning using the raster scan method consists of a horizontal periodic signal 201 and a video signal 202. It is expressed as a value. and a maximum brightness level of 20, which corresponds to the brightest part of the image.
3 and the blank level 204 corresponding to the darkest part is 0.7■. This video (No. A) is converted into a 256-gradation digital signal by, for example, an 8-pitch A/[) converter;
This is done by dividing the area between the maximum brightness level 203 and the black level 204 into 256 equal parts, and making the small ranges of the equally divided levels correspond to digital values 0 to 255 in order from the bottom. Therefore, the maximum brightness level 203 is converted to a digital signal value 255, the black level 204 is converted to a digital signal value O, and when the level is an intermediate level, it is converted to a digital signal value corresponding to a small range including that level.

When digitizing video signals in this way, there are the following problems. Assume that the screen to be digitized has a constant brightness portion 205 as shown in FIG. Naturally, when sampling and digitizing this, a sampling pulse is used. However, this pulse contains many harmonic components and is applied to a circuit close to the Video Zeo No. 44 202 to be sampled, so even if measures are taken for one circuit, a noise voltage 206 of an integral multiple of the sampling period will be generated. Fourth
It is unavoidable to add weight to the video signal as shown in the figure.

This noise voltage is sufficiently small considering the width of the small range of levels divided by 256, so the military level 205 is the fourth level.
As shown in the figure, when it is near the middle of the threshold levels 402 and 403 (the small range corresponding to the digital value 73) indicating the boundary of the small range (digital value 73), the digitized signal value remains 73. Therefore, the noise 206 does not pose any problem. In other words, the flat level 205 as shown in Figure 4 continues for example for 9 sampling periods,
Moreover, if the same level continues in the vertical direction,
The sampling values of this flat portion are all 73 as shown in FIG. However, even with similar flat images,
As shown in Figure 6, the flat level is the threshold level 40.
2, the superimposed part of the nozzle voltage 206 falls within a small range one level above, so the digital value at this point becomes 74, and as shown in FIG.
A level of 13L is applied to the digitized signal. This vertical stripe noise does not become a problem if the screen brightness is not flat and changes significantly, and no special countermeasures have been taken in the past, but in recent years, high-quality and faithful digitalization It is necessary for image analysis, etc.

In addition, there is a filtering process as a means of removing noise from images, and the general content is disclosed in ``Computer Image Processing'' (Takeshi Yasuoin, ``Masayuki January Nari, Sanpo Shuppan. 197
Released in June 2009. There are IF¥s after 15 pages. However, sampling noise is not taken into account.

[Purpose of the invention]

An object of the present invention is to provide an effective vertical stripe noise removal device for removing the above-mentioned vertical stripe noise caused by sampling pulses.

[Summary of the invention]

The present invention is characterized in that a pixel in which vertical stripe noise occurs is extracted by a convolution calculation, and only the signal level corresponding to this pixel is corrected according to another convolution calculation.

[Embodiments of the invention]

The present invention will be explained in detail below. FIG. 1 shows an embodiment of the present invention, in which the fifth and fifth digital values of the first raster of a digitized screen 1 are assumed to be f l,J. f□ and J have values as shown in FIG. 5 and FIG. 7, for example. Distributor 2 has one line (one row of molecules in the horizontal direction of screen 1, l+1+ft+z).

f11. n is the number of samples for one row in the horizontal direction = number of pixels) It is composed of a delay circuit, etc., and during the correction operation of f1.1, L, J-L f+, J f++
++1 ・(])f I"l+ J-
1f + "1 + Jf t" 11 J" 1 is taken out and set in data unit 3. In data unit 3, in order to perform a convolution operation corresponding to the set position of each data in (1) 9 coefficients are preset, and 9 processor elements 1
-4 is multiplication ゛Wk, *x f l"JJ"l, l < k, Q
< Execute 1-(3). The combination unit 5 adds all the values of equation (3) calculated by each processor element 4 to obtain the value. After passing through the absolute value circuit 6, this output RIJ is binarized by the binarization circuit 7 when vertical stripe noise is detected and sent to the binary screen 8-1 as mask data, and when the screen is rewritten, it is directly sent to the screen 9. They are written as the th and 5th values of the jth rask, respectively.

Here, a dedicated memory (or memory area) is required for the binary screen 8, but the screen 9 and the screen 1 after rewriting may be physically separate from each other in memory.

It is also possible to use an appropriate register to create the same memory as the screen], and part b can be determined in relation to the purpose of use of the modified screen and the device cost (the reason why registers are necessary when making the screen the same is as follows f l-1 before correction in calculating gIi
+ J-1f 1-1n J f I-1! J”l
This is because data from the previous line is required).

Next, the operation of this embodiment will be explained. Assuming that the data of meeting ceremony (1) is stored in data unit 3 from cell 1 to cell 1, and each data represents a portion of the screen at flatness level X as shown in Fig. 6, when there is no vertical stripe noise, the formula ( The values of each f in 1) are all X. However, the point that is currently being revised is f+a.
Assuming that vertical stripe noise is included at , Equation (1) becomes However, the digital value is an integer from 0 to 255, and the sampling pulse noise 206 (Figure 4.6)
is assumed to be sufficiently small, so it is an integer between X-0 and 255, α
=+1 or -1. Therefore, the coefficient Ws+a of equation (2)
If we write gIJ calculated by equation (4) when is determined as hIJ, it can be easily seen from equations (5) and (6) that h-th = α. This result means that the vertical stripe noise component α at the position of data flJ can be detected by performing the convolution calculation of equation (4) using the coefficient of equation (6)1. Therefore, in this embodiment, the coefficient W□ of equation (6) is set in advance in the data unit 3, and this is first used to execute the convolution operation of equation (4).

Extract the vertical stripe noise α. At this time, the absolute value of the extracted noise α is calculated by the absolute value circuit 6 in FIG. 1, and the result 1α1 is binarized by the binarization circuit 7. The characteristics of this dichotomous circuit 7 are shown in FIG. 8, and if the input 1αl is between 0.5 and 1.5, it outputs 1, and otherwise it outputs O. 6 As mentioned above, In the case of vertical stripe noise caused by the influence of sampling pulses,
Since 1α1=1, the output of the binary northbound IM7 is 1 when the vertical stripe noise α is included in the digital value flJ to be corrected, and O otherwise. The output of the binarization circuit 7 is written as the j-th data value of the j-th rask on the binary screen 8, and is used as mask data when the screen 1 is rewritten later.

Convolution operation using the following coefficient Wht (4)
If we write g l J tc' l-1t 7 when executing, (1) and (8) it is easy to obtain HIi= f IJ= X + a
-(D). Therefore, from equations (7) and (9), G+1=HIJ h+1=(x+α) a=x
- (In) In other words, the correct value g+J after the correction is obtained by removing the noise component α. However, the convolution operation (4) is linear with respect to "f+", and Equation 1 (10)
uses equations ((3) and (8)) [Since both require convolution for flJ in equation (1),
The difference between HI J and h+a in equation (10) is the coefficient (8) and (6
) is determined by convolution of f+a using the difference between them as a coefficient. In other words, the coefficient obtained by subtracting each coefficient of equation (6) from each coefficient of book (8) is 636 n 1] - 1 636, □. Therefore, using the coefficient of equation (11) as wkffi, calculation (4) is performed as fll. , G15=x in equation (10) can be obtained with one convolution operation. Therefore, the coefficients of this equation (11) are also set in the data unit 3 in advance. After generating the binary screen 8 using the coefficients of equation (6), this time we use the coefficients of equation (11) to
When the convolution operation is performed, Equation (10)
Since G I J is obtained, when the j-th value of the corresponding first rask on the binary screen 8 is 1, rewrite f 1j on the screen]- with the G I J obtained here. By doing this, it is possible to obtain the screen 9 from which vertical stripe noise has been removed. When the corresponding value on binary screen 8 is O, it is assumed that there is no vertical stripe noise, and screen 1 is not replaced (
i.e. mask). In addition, in FIG. 1, it is assumed that GIJ is also input to the screen 9 after passing through the absolute value circuit 6.
Since G I J is a positive value of 0 to 255, it is the same value even if it is not passed.

〔Effect of the invention〕

As is clear from the examples below, according to the present invention,
This has the effect of reliably removing vertical stripe noise on a flat screen.

[Brief explanation of drawings]

Fig. 1 is a diagram showing an embodiment of the present invention, Fig. 2 is an explanatory diagram of the video [No. 1], and Figs. 3 to 7 are illustrations of the video
FIG. 8 is an explanatory diagram of vertical stripe noise that occurs when digitizing. FIG. 8 is a characteristic diagram of a binarization circuit. 1.9-=stroke jn1.2 minute W device, 3--F-9:L
=′y+゛.・1 ・Processor element, 5... Combiner, 6...
Absolute value circuit, 7. Binarization circuit, 8. Binary screen.

Claims (1)

[Claims] 1. In a noise removal method for removing vertical stripe noise superimposed on a digitized digital screen signal, a digital signal of a pixel in a neighboring area centered on a pixel to be corrected at a certain point in time is used. calculating the product of a distribution means for taking out the digital signal, a coefficient storage means for storing in advance coefficients corresponding to each of the taken out digital signals, and the above-mentioned coefficient corresponding to each of the digital signals taken out by the said distribution means. Then, a convolution operation means for calculating the sum of the calculated products is provided, and when the digital signal values in the vertical and horizontal directions of the pixels in the neighboring area taken out by the distribution means do not change linearly, the straight line is calculated. The first coefficient group is determined to detect a shift based on a change in the shape, and the first coefficient group is determined to extract only the digital signal of the correction target pixel from the neighboring region pixel. A second coefficient group generated by subtracting each corresponding coefficient of the coefficient group is preset in the coefficient storage means, and an output of the convolution calculation means calculated using the first coefficient group. When the absolute value of is within a predetermined range, and only then, the digital signal of the correction target pixel is rewritten with the output of the convolution calculation means calculated using the second coefficient group. A device for removing vertical stripe noise, characterized in that: 2. The vertical stripe noise removing device according to claim 1, wherein the neighboring area pixels are 3×3.