SU1305735A1 - Device for correcting television pictures - Google Patents

Abstract

The invention relates to technical cybernetics and can be used in automatic image recognition systems, for example, to recognize printed or handwritten characters, fingerprints, drawings and other objects. The purpose of the invention is to increase the noise immunity of a device. The image correction device allows for image segmentation and contains image sensor, analog-to-digital converter, digital median filter, count delay elements (registers), edge detector, discharged from the subtractors, multipliers at a constant rate, comparators and the switch, a discriminator and a synchronizer. The device is distinguished by an increased reliability of image segmentation due to the possibility of enlarging the edges of the image line invariant to the magnitude of the difference in brightness. The introduction of a digital median filter improves the noise immunity of the device, while the segmentation accuracy is not reduced, since the median filter does not smooth out sharp differences in brightness. Parallel to the pipeline, the organization of computations in the device allows the segmentation of television images in real time. 1 z.p, f-ly, 4 ill. iS (/ OS ate with SP

Description

The invention relates to the field of technical cybernetics and to be used in automatic image recognition systems, for example for recognizing printed or handwritten characters, fingerprints, drawings and other objects that can be uniquely represented by binary images.

Celb invention - improving the noise immunity of the device.

Fig, 1 shows a block diagram of the device; FIG. 2 is a block diagram of a digital median filter, which is equal to three; on fig.Z - the discriminator bioc scheme; 4 shows diagrams explaining the method of adaptive line segmentation of images, which is implemented in the device.

to

15

a sharp drop in brightness. Then one of the two conditions for the presence of a sharp drop in bone at the current point is performed by comparing the steepness (sharpness of the drop) at a given point with the maximum permissible edge. If the current point is on the border of the background object, the result of the segmentation is this point is 1, otherwise - O,

The device works as follows.

Image samples in the course of its line scanning by the image sensor 1 are successively input to the ACP 2. B ADC 2 converts the digital image into the image in accordance with the hour 25

The device contains a sensor of such a clock synchro pulses at the output of the clock, an analog-digital re-processor 14, a clock frequency (ADC) 2, a digital median filter 3 p delay elements 4, 5 edge detector, consisting of the first 6 and the second 7 drivers, the first 8 and second 9 multipliers, the first 10 and second 11 comparators and the switch (multiplexer) 12, the discriminator 13 and the synchronizer 14,

Digital median filter (figure 2) contains the first and second delay elements 15 delay, two comparator 16, node

17 memory, multiplexer

18 and the third delay element 19, the Discriminator 13 (FIG. 3) contains

woven with the help of two elements of 15 delays, in turn, at the exit

two adders 20, two subtractors 21 and a comparator 22,

The combination is chosen so that the distance between two readings does not exceed half the length of the string for the shortest possible object. Median filter 3 implements the removal of impulse noise in the image line, which can be mistaken for objects, t, e, at these points there can be sharp differences in brightness. By means of two comparators 16, the current image reference g (t;) at the output of the A / D converter 2 is compared with two previous g (t, v,) and g (t ;.) zad30

45

The third delay element 19 for the current device operation cycle will be a method of line-by-line segmentation of the image; 40 is the result of comparing samples g (ti,)

and g (t,) calculated in. previous cycle through the first comparator 1 6 p

Based on the results of pairwise comparison of three consecutive samples of the image line g (t) using the constant node 17, the table contains the number of that of the three samples g (t;), g (t,) and g (t ,.) which is the median. For this, the address input of node 17 is connected to the outputs of two comparators 16 and trigger 9, and its information output j is connected to the control input of multiplexer 18. Thus, in each operation cycle of the device at the output of multiplexer 18, i.e. at the output of the digital median filter 3, on the proposed device implements

Research institutes of the input image on an object and background by adaptive and noise-resistant detection of object-background borders, i.e. sharp drops in brightness. Preliminary, in order to improve the noise immunity of the device, digital median filtering of the input image is performed. Median filtering has the advantage over linear filtering that it allows you to remove (smooth out) interference and not smooth out sudden drops in brightness. To detect the boundaries of the object-background, t, e. sharp drops in brightness, it is determined what side of the edge is the current point of the image line, provided that at this point there is

50

five

a sharp drop in brightness. Then one of the two conditions for the presence of a sharp drop in bone at the current point is performed by comparing the steepness (sharpness of the drop) at a given point with the maximum permissible edge. If the current point is on the border of the background object, the result of the segmentation is this point is 1, otherwise - O,

The device works as follows.

Image samples in the course of its line scanning by the image sensor 1 are successively input to the ACP 2. B ADC 2 converts the digital image into the image in accordance with the hour

clock sync pulse output 14, clock frequency

The combination is chosen so that the distance between the two readings does not exceed half the length of the string for the minimum possible object. Median filter 3 implements the removal of impulse noise in the image line, which can be mistaken for objects, t, e, at these points there can be sharp differences in brightness. By means of two comparators 16, a comparison of the current image reference g (t;) at the output of the A / D converter 2 with the two previous g (t, v,) and g (t ;.) is performed,

The result of the median filtering f (t) is calculated. FIG. 4a shows a fragment of the image line at the input of the ADC, and FIG. 4b is the result of the median filterdia of this line.

The discriminator 13 is used to determine on which side of the edge the current point of the image line f (t) is located, provided that there is a sharp difference in brightness at that point. The discriminator 13 calculates the value of the variable a- according to the following formula

1, (/ f (t;) + f (t ;,) -f (t, -2 -) / - / f (t; M) + f (t; .2) -f (t;) /) sO;

Oh, otherwise.

If a is 1, then the count fCt ;;.,) Is on the right side of the boundary; if a, 0, then it is located on the left side. The variable a corresponds to the output signal of the comparator 22, the inputs of which receive the absolute values of the numbers at the outputs of the switches 21, for example, in FIG. 4b the count f (t;.,) Is located on the right side of the border.

With the help of the 5 edge detector, sharp differences in the brightness of the image line at the output of the median filter 3 are detected. At the same time, the output signal a is learned; the discriminator 13, which is fed to the control input of the multiplexer 12. If the output signal of the discriminator 13 is zero, then the output signal L of the edge detector 5 is determined as follows,

1. / f (t;.,) - f (t; -i) /. Ci / f (t, O

- f (ti-,);

B; one

Oh, otherwise.

where o /, is the specified marginal sharpness coefficient of the edge, which is constant for all rows of the specified images.

The value of bj is calculated by means of the first 6 and second 7 subtractors, multiplier 8 and comparator 10. Since the sharpness coefficient of edge 6 is constant, in order to increase the speed of the device, both multiplier 8 and multiplier 9 are implemented as a permanent memory (ROM) , to the address inputs of which the output signals of the subtractors are received, i, e. the multiply operation completes the table. If the output signal of the discriminator 13 is single, then the output signal of the edge detector 5 is determined as follows

B.

l, / f (ti) -f (t; -,) (t ;,) -f (ti.2);

Oh otherwise

those. the output of the detector 5 edge receives the output signal of the comparator 11. The drive signals of the subtractors 6 and 7 do not contain the sign bit of the result of the subtraction, t, e. they correspond to the absolute value of the difference of two numbers. Thus, as a result of the operation of the device, segments of the image lines with different brightness levels will be selected, the ends of which correspond to single values of the variable b | For example, on

fig 4b point t.

will be allocated

five

0

five

0

5 I

as a point of a sharp drop in brightness, and a point as a point of smooth change in brightness, i.e., a point is unlimited.

The value of the variable aj indicates whether the detected abrupt drop point is the beginning of a new segment or an end. In case the edge was not detected with the 5 edge detector, then the value of the variable a; determines the increase or decrease of the luminance of the image line relative to three adjacent points. The calculations in the edge detector 5 and the discriminator 13 are carried out simultaneously (in parallel), these units operate in a conveyor mode. Such a parallel-pipeline mode of operation of the device significantly increases its speed.

0

five

Thus, due to the possibility of separating the edges of the image line invariant to the magnitude of the difference in brightness, the proposed device has an increased quality (reliability) of image segmentation, since only the marginal sharpness coefficient is specified. The introduction of a median filter allows to increase the noise immunity of the device and at the same time does not reduce the segmentation reliability, since the median filter does not smooth out sharp differences in brightness. Parallel conveyor organization

Immersion in the device makes it possible to segmentation of television images in real time.

Claims (1)

Invention Formula 1, A television image correction device comprising a TV ion image sensor, whose output is connected to the input of an analog-digital converter, a synchronizer, a first and second subtractors, a first and second comparators, a first multiplier, a switch, the output of which is an information output of the device, and the first and second delayed elements are connected in series, the output of the first subtraction — Tel is connected to the input of the first multiplier, and the output of the synchronizer is connected. not with synchronizing inputs of analog-digital converter, delay elements, subtractors of the first multiplier and comparators, which is so that, in order to increase the noise immunity of the device, a digital median filter, a second multiplier and a discriminator, whose information inputs are connected, are introduced into it respectively, to the outputs of the digital median filter, the first and second delay elements, and the output is connected to the control input of the switch and is the control output of the device; the output is analog-to-digital th transducer is connected to the input of a digital median filter whose output is connected to the input of the first element Delays and a first input of the first You are a reader, whose output is connected to the first input of the second comparator, a second input coupled to vyho the house of the second multiplier, the output of the primary memory node 45, The second delay element is connected to the second input of the first subtractor and to the first input of the second subtractor, the second input of which is connected to the output of the second delay element, and the output is connected to the input of the second multiplier and another input of the first comparator j another input connected to the output of the first multiplier j the outputs of the first and second comparators are connected respectively to the first and second information inputs of the switch; and the synchronizer output is connected to sync; inputs of a digital median filter, a second multiplier and a discriminator, 2 i The device according to p., About tl and - the fact that in it the digital median filter contains three delay elements, two comparators; , a fixed memory node and a multiplexer p whose output is the output of a block whose input is connected to the input of the first delay element, the first information input of the multiplexer and the first inputs of the first and second comparators 5 whose outputs are connected to the first and second address inputs of a constant node the memory, the third informational input of which is connected to the output of the third delay element, whose input is connected to the output of the first comparator, the outputs of the first and second delay elements are connected respectively to the second and third information multiplexer and the second input1 "g of the first and second comparators 5 outputs of the node this memory is connected respectively to the first and second control inputs of the multiplexer, the synchronizing input of the block is connected to the synchronization inputs of the delay elements and (Rig.1 Fig.Z 6 Sin) (p. 4-2 ii l t k-2 tf (-j tff Editor V.Danko Compiled by T.Nichiporovich Tehred L. Serdyukova Order 1455/49 Circulation 673. Subscription VNIIPI USSR State Committee for Inventions and Discoveries 113035, Moscow, Zh-35, Raushsk nab ,, d, 4/5  - - ------ - "--- e-fc. --------- ..-.-.- I-. , "..," ",. and .. --- .- .-- c. . ,, „.,., - Production and printing company, Uzhgorod, Projecto st., 4 FIG Proofreader E.Roshko