SU1320876A1 - Two-dimensional digital filter - Google Patents

Description

one

The invention relates to radio engineering and can be used in digital image signal processing systems.

The purpose of the invention is to improve the accuracy of filtration.

Figure 1 shows an electrical block diagram of a two-dimensional digital filter; Fig. 2 is the same, the dispersion calculation unit; Fig. 3 is the same, block averaging; 4, the non-block of the formation of the selective median; fig. 3 - the same, comparator module.

A two-dimensional digital filter (FIG. 1) contains analog video signal of sensor I, analog-digital converter 2, p-1 delay elements 3-1-3-P-1 for counting, first block 4

1320876

.ten

five

signal in real time in order to reduce the level of tea noise. The value result of filtering rows or columns is matrix. The discrete images for each sample are calculated depending on the local properties of the video signal. For this, for each image sample, the image dispersion is calculated by its n consecutive samples (rows, and then columns) and compared with a predetermined constant threshold, which is previously recorded in the threshold memory block 10. If the variance value exceeds the threshold, then median filtering is performed, which, together with noise removal, maintains sharp differences in brightness. Median filtration also

forming a selective median, first well removes impulse noise, a second group of n multipliers 5-1-5-n. If the variance is less than the specified semi-adder 6, first block 7 of the coefficient memory, first block 8 of the variance estimate, first switch 9, threshold memory block 10, p-1 delay elements 1 1-1-П-П-1 per line, second block 12 forming a selective median, second group of n multipliers 13-1-13-p, second adder 14, second block 15 memory coefficients, the second unit 16 dispersion estimates, the second switch 17, the generator 18 clock pulses.

The first and second blocks 8 and 16 of the dispersion estimate (Fig. 2) contain the first AND second blocks 19-1, 19-2 averaging, the first and second squares 20-1-20-2, block 21 of the subtraction, block 22, the comparison, trigger 23.

The first and second blocks 19-1, 19-2 averaging (Fig. 3) contain a delay element 24, a subtractor 25 / accumulator adder 26, a divider 27.

The first and second blocks 4 and 12 of the formation of the selective median (Fig, 4) contain n blocks 28-1-28-p

permutations of numbers, each of which consists of (p-1) / 2 comparator modules 29-1-29- (p-1) / 2 and a delay element 30,

The comparator module (Fig. 5) contains a comparison unit 31, the first and second switches 32-1, 32-2.

45 From the sensor 1, the image in the process of line scanning it is fed to the input of analog-digital converter 2, which converts the input video signal into digital form.

50 From the information output of the A / D converter 2 and from the outputs of p-1 delay elements 3-1-З-p-1, the image lines in parallel are fed to the corresponding input two-dimensional digital filter operation-55 of the first block 4 forming the selective median and to the first inputs corresponding multipliers 5-1-5-p of the first group; In the first block 4, the formation of the selective median is as follows.

A two-dimensional digital filter performs adaptive filtering of analog, for example, television video. 1320876

0

five

signal in real time in order to reduce the level of tea noise. The value result of filtering rows or columns is matrix. The discrete images for each sample are calculated depending on the local properties of the video signal. For this, for each image sample, the image dispersion is calculated by its n consecutive samples (rows, and then columns) and compared with a predetermined constant threshold, which is previously recorded in the threshold memory block 10. If the variance value exceeds the threshold, then median filtering is performed, which, together with noise removal, maintains sharp differences in brightness. Median filtration also

five

0

then linear video filtering is performed in digital form. The impulse response coefficients of the linear filter are pre-recorded in the first and second blocks 7 and 15 of the coefficient memory.

If a threshold value equal to zero is preliminarily recorded in threshold memory block 10, the dispersion will always exceed this threshold value and the proposed two-dimensional digital filter will perform a separate (first by row and then by 5 columns) median image filtering. If the maximum allowable number is written into the memory block 10 of poro.ca, then the dispersion for all image samples will always be less than

This threshold value will be separated into a linear image filtering, regardless of its local properties.

0

From the sensor 1, the image in the process of line scanning it is fed to the input of analog-digital converter 2, which converts the input video signal into digital form.

From the information output of A / D converters 2 and from outputs n-1 of delay elements 3-1 to 3-n-1, parallel-conveyor exchange sorting of a sequence of n samples of rows is parallelized in counting n consecutive reports. By means of the first block 8, the variance estimate in each operation cycle of the device the variance is calculated from n counts within the image line. The value

d j i ix4-ui -fi-ixk-i: if

. in, -. J In u, J The squaring operations are implemented by the first and second quadrants 20-1, 20-2 tabulated using ROM. To calculate the average value of n counts within a line, use the first averaging block 19-1, and to calculate the average value of squares A second block 19-2 averaging is applied to the samples within the row. Then, in each filter cycle, the variance estimate d is generated at the output of the subtraction unit 21. In the first and second blocks of averaging 19-1, 19-2, the average value is calculated recursively, for example, in the first block 19-1 averaging X | is calculated based on the following relationship:

 to

where X is the count of the image line at its input; X c., Is the countdown at the output of delay element 24, which delays by n counts. The division by p operation is implemented in the divider 27. In the comparison block 22, the variance estimate is compared with the threshold value stored in the memory block 10. If the dispersion for the current signal reading exceeds the threshold value, then the trigger 23 is set to one, In this case,

g (x.'s, ..

),

The one-dimensional median filter is placed in the first and second adders whose inputs

input line strings

Otherwise,

linear filtering of input lines

images,

From the output of the first switch 9, samples of the result of line adaptive filtering of the image are fed to the input of the first delay element 11-1 per line and the second block .16 for50

connected, respectively, to the outputs of the first and second groups of p multipliers, which is similar to the fact that, in order to increase the filtration accuracy, the first median unit for the formation of a selective median, the inputs of which are connected to the inputs of the delay elements per count, and the first switch are introduced, second

Worldwide selective median. The second 55 signal input of which is connected to

block 16 generates a control output of the first adder, a post-signal depending on the locally connected second block

properties of the columns of the matrix of samples forming a median,

images, in the first and second block inputs of which are connected to the inputs

four

In averaging the second dispersion estimator 16, a delay element 24 is used, delaying n lines of the image. Through the second block 12 of the formation of the sample median, p multipliers 13-1-13 p of the second group, the second adder 14, the second dispersion estimator 16 and the second switch 17

g, adaptive filtering is performed on the columns of the matrix of image samples at the output of the first switch 9, which is similar to the described adaptive line filtering. Thus, in each cycle of operation of a two-dimensional digital filter, at the output of the second switch 17, the next sample of the result of a two-dimensional digital filtering of the input image is formed,

0

five

0

five

0

F o rmula of the invention

Claims (2)

1, A two-dimensional digital filter comprising an analog video signal sensor connected in series, an analog-to-digital converter and n-1 serially connected delay elements per sample, a first coefficient memory block and a first group of n multipliers, the first inputs of which are connected to the corresponding outputs of the first memory block coefficients, and the second inputs are combined with the inputs of like delay elements on the countdown, n-1 serially connected delay elements on the line, the second block of memory coefficients, and the second group of p multipliers, the first inputs of which are connected to the corresponding outputs of the second coefficient memory block, and the second inputs are combined with the inputs of the same name delay elements per line, as well as 0 connected, respectively, to the outputs of the first and second groups of p multipliers, which is similar to the fact that, in order to increase the filtration accuracy, the first median unit for the formation of a selective median, the inputs of which are connected to the inputs of the delay elements per count, and the first switch are introduced, second delay elements for the row and the second switch, the second signal input of which is connected to the output of the second adder, as well as the threshold memory block and the first and second dispersion estimation blocks, the first inputs of which are combined and connected to the output of the threshold memory block, the second outputs The connectors are, respectively, with the output of the analog-digital converter and the first switch, and the outputs are connected with the control inputs of the first and second switches, respectively. 2. The filter according to claim 1, characterized in that the first and second fie. 2 Entry about J five The dispersion estimator blocks are made in the form of serially connected first quad, first averaging block, subtraction block, comparison block, the other input of which is the first input of the dispersion estimator, and a trigger whose output is the output of the dispersion estimator, as well as serially connected second block the averaging, the input of which is combined with the input of the first quadrant and is the second input of the dispersion estimator, and the second quadrator, the output of which is connected to another input of the block. subtraction. Output Out Sync o (pae.S Editor N.Gorvat Compiled by E. Borisov Tehred A. Kravchuk Order 2666/55 Circulation 901 Subscription VNIIPI USSR State Committee on affairs of inventions and discoveries II3035, Moscow, Zh-35, Raushsk nab, 4/5 Production and printing company, Uzhgorod, Projecto st., A Vlodsin fofie. four Proofreader V. But ha