JPH04276887A - Image processor - Google Patents

Abstract

PURPOSE:To make it possible to accurately carry out measuring feature quantity and pattern matching by emphasizing edge to eliminate the effect of illumination fluctuation and realizing preprocessing where noise component is not emphasized. CONSTITUTION:The present image processor 40 includes a preprocessing circuit 42 constituted of an A/D converter 47, a differentiating arithmetic section 48, a multiplier 49, a divider 50, and a adder 51. In the differentiating arithmetic section 48 a gradation image signal of digital variable is inputted, and the first order differentiating arithmetic section 52 calculates an absolute value of the first order differentiation and the second order differentiating arithmetic section 53 the second order differentiated value, respectively. The multiplier 49 carries out multiplication of each differentiation arithmetic output, the divider 50 divides a multiplication output by a constant, and the adder 51 adds the division output to the gradation image signal, thereby generating an output signal to a binarization circuit 43.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] This invention binarizes a grayscale image signal obtained by imaging an object, and uses the binarized image to perform various processes such as feature measurement and pattern matching. The present invention relates to an image processing device, and particularly relates to a preprocessing circuit that performs predetermined preprocessing on a grayscale image signal before binarizing the grayscale image signal and outputs the preprocessed signal to a binarization circuit.

0002

2. Description of the Related Art In general, an image processing device converts an analog grayscale image signal obtained by imaging an object using an imaging device into a digital grayscale image signal, and then converts the digital grayscale image signal into a fixed grayscale image signal. Binarization processing is performed using a binarization threshold to generate a binary image signal.

[0003] FIG. 14 shows a gray scale image 1 obtained by imaging the letter "A" under predetermined illumination. It shows.

FIG. 15 shows a brightness distribution 5 along the horizontal line 4 of the grayscale image 1. In the figure, the parts indicated by 5a and 5b are the boundary parts between the characters and the background (hereinafter referred to as "edge parts").
The brightness distribution at this edge portion is gently sloped due to the influence of the optical system of the imaging device and the state of the object.

Further, in FIG. 15, 6 shows the brightness distribution when the imaging conditions change due to illumination deterioration, and the brightness level is lower overall compared to the brightness distribution 5 under proper illumination. .

[0006]

[Problem to be Solved by the Invention] When each grayscale image is binarized using the binarization threshold value TH shown in FIG.
When the illumination is degraded, the width d of the "0" level is greatly different, as shown by the reference numeral 8.

FIG. 16 shows a binary image 9 under proper illumination and a binary image 10 under degraded illumination in comparison. In the figure, the blacked-out part shows a black pixel area representing the character image part 11, and the other parts show the white pixel area representing the background image part 12, respectively. According to the figure, in the binary image 10 when illumination has deteriorated, the character image portion 11 is extremely thick compared to the binary image 9 under proper illumination. Therefore, when a predetermined window 13 is set in these binary images 9 and 10 and the area of the black pixel area within the window 13 is measured to perform shape recognition, etc., the measured value of the area changes when the illumination deteriorates. This results in erroneous recognition.

FIG. 17 shows a standard pattern 14 obtained by capturing an image of the letter "A" under proper illumination, and a binary image 15 obtained by capturing an image of a recognition object under degraded illumination. This binary image 15 is an image for the letter "A", but when pattern matching is performed by comparing this binary image 15 with the standard pattern 14, as shown in the matching image 16 in the same figure, a mismatch occurs. As a result, pixels (blacked out areas in the figure) occur over a wide range, making proper recognition difficult.

The above-mentioned problem is caused by the gradual slope of the brightness distribution at the edge portion of the grayscale image.
To solve this problem, it is also possible to adopt a method of edge-emphasizing the edge portions of the grayscale image using a Laplacian operator or the like.

However, when this method is used, if a large number of noise components are present in the grayscale image, edge enhancement will result in noise components being emphasized in addition to the edge portions. For this reason, when area measurement or pattern matching is performed using a binary image obtained from a grayscale image after edge enhancement, the measured area value changes due to the influence of noise components, and the number of mismatched pixels increases. Similarly, this results in erroneous recognition.

FIG. 18 shows a grayscale image 17 obtained by imaging the letter "A" under predetermined illumination and performing edge enhancement. In the figure, 18 is the image portion of the letter, and 19 is the background Each image part is shown. In this grayscale image 17, a large number of noise components 20 appear in each image portion 18, 19 of characters and background.

FIG. 19 shows the horizontal line 21 of the grayscale image 17.
A brightness distribution 22 along the . This brightness distribution 2
2 with the brightness distribution 5 shown in FIG. 15, the slopes of the edge portions 22a and 22b are steep.

Further, in FIG. 19, 23 indicates the brightness distribution when the imaging conditions change due to illumination deterioration.
The brightness level is lower overall than the brightness distribution 22 under proper illumination.

When each grayscale image is binarized using the binarization threshold TH, the binarization level along the horizontal line 21 becomes 24 under proper illumination, and 25 when the illumination is degraded. The width d of the "0" level is almost the same, but since the noise component is also emphasized by edge emphasis, noise parts 26 and 27 are everywhere at the binarized level.
In addition, they appear in different positions and in different sizes under proper illumination and when illumination has degraded.

FIG. 20 shows a binary image 28 under proper illumination and a binary image 29 under degraded illumination in comparison. In the figure, the blacked out area is the black pixel area representing the character image area 30, the other area is the white pixel area representing the background image area 31, and the spots are the noise image areas 32, 3.
Pixel areas representing 3 are shown respectively. These binary images 2
When a predetermined window 34 is set in the windows 8 and 29 and the area of the black pixel area within the window 34 is measured to perform shape recognition, etc., the binary image 29 under illumination degradation and the binary image under proper illumination are used. The value image 28 is the noise image portions 32 and 33.
Since the positions and areas of the two do not match, the measured value of the area changes when the illumination deteriorates, resulting in erroneous recognition.

FIG. 21 shows a standard pattern 35 obtained by capturing an image of the letter "A" under proper illumination, and a binary image 36 obtained by capturing an image of a recognition target under degraded illumination. This binary image 36 is an image for the letter "A", but when pattern matching is performed by comparing this binary image 36 with the standard pattern 35, the position is changed as shown in the matching image 37 of the same figure. and mismatched pixels (in the figure,
As a result, proper recognition is difficult.

[0017] The present invention was made with attention to the above problem, and it eliminates the influence of illumination fluctuations by edge emphasis, and also realizes preprocessing that does not emphasize noise components, thereby making it possible to accurately measure feature quantities and pattern matching. The object of the present invention is to provide an image processing device that can perform the following operations.

[0018]

[Means for Solving the Problems] The image processing device of the present invention includes a preprocessing circuit that inputs a grayscale image signal obtained by imaging an object and performs preprocessing, and a binary output signal of the preprocessing circuit. and a binarization circuit that performs conversion processing to generate a binary image signal,
The preprocessing circuit is provided with a differential calculation section, a multiplication section, and an addition section. The differential calculation unit inputs the grayscale image signal, calculates the absolute value of the first-order differential and the second-order differential, and outputs a first-order differential signal and a second-order differential signal. The multiplication section receives the first-order differential signal and the second-order differential signal from the differential operation section and performs multiplication processing. The adder adds the multiplication output signal from the multiplier to the grayscale image signal to generate an output signal to the binarization circuit.

[0019]

[Operation] Since the product of the absolute value of the first-order differential of the gray-scale image signal calculated by the differential calculation unit and the second-order differential value is added to the gray-scale image signal, the second-order differential value The absolute value of the first-order differential functions to prevent the noise portion from being emphasized. As a result, preprocessing is achieved in which the edge portions of the grayscale image are emphasized but the noise components are not. This eliminates the influence of illumination fluctuations and allows accurate measurement of feature quantities and pattern matching.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a schematic configuration of an image processing system in which the present invention is implemented, and includes a camera 41 and an image processing unit 40. The camera 41 images an object (for example, characters) under illumination by an illumination device (not shown), and outputs an analog grayscale image signal VDi to the image processing device 40.

The image processing device 40 includes a preprocessing circuit 42
The preprocessing circuit 42 includes an A/D converter 47, a differential operation section 48, a multiplier 49, and a divider 50. , an adder 51, and the like.

The preprocessing circuit 42 processes the grayscale image signal VD.
The A/D converter 47 inputs the analog grayscale image signal VDi from the camera 41, converts it from analog to digital, and converts it into a digital signal. The grayscale image signal VD of the amount is outputted to the differential calculation section 48 and the adder 51.

The differential calculation section 48 is a first-order differential calculation section 52.
and a second-order differential calculation section 53. The first-order differential calculation section 52 inputs the grayscale image signal VD, calculates the absolute value of the first-order differential, and outputs the first-order differential signal J to the multiplier 49. Further, the second-order differential calculation section 53 inputs the grayscale image signal VD, calculates a second-order differential value, and outputs the second-order differential signal K to the multiplier 49 .

The second-order differential calculation section 53 is for emphasizing the edge portion of the grayscale image, and the first-order differential calculation section 52 is for preventing noise components from being emphasized due to edge emphasis. 4 shows the principle of the first-order differential calculation section 52 and the second-order differential calculation section 53.

In FIG. 4, reference numeral 61 indicates a grayscale image input to the differential calculation section 48. A rectangular mask 62 is set on this grayscale image 61, and the entire image is scanned in the direction of the arrow. Each square in this mask 62 corresponds to one pixel, and the pixel data of the pixel of interest centered on e is stored in a to d and f.
~i indicates pixel data of pixels in eight surrounding directions, respectively.

A Sobel operator that is resistant to noise is used as the first-order differential calculation section 52, and the absolute value of the first-order differential is calculated by performing the calculation according to the following formula (2) at each pixel position while scanning the mask 62. do. In the case of using this Sobel operator, as is clear from equation (2), the absolute value of the difference between the pixel data of each vertically and horizontally adjacent pixel to the center pixel of interest is calculated and added, so for example, spike noise may exist in the pixel of interest. Even if it is, it has no effect on the absolute value of the first derivative.

[0027]

[Math 1]

A Laplacian operator is used as the second-order differential calculation unit 53, and while scanning the mask 62, a calculation according to the following equation (2) is executed at each pixel position to calculate a second-order differential value.

[0029]

[Math 2]

FIG. 5 shows a density distribution 63 along a predetermined horizontal line for a certain grayscale image. In the figure, 63a is an object image, 63b is a background image, and 63c is
correspond to the edge parts of the object, respectively.

FIG. 6 shows the calculation output when the grayscale image signal having the density distribution 63 shown in FIG. It protrudes in the positive direction corresponding to the position of.

FIG. 7 shows the calculation output when the grayscale image signal having the density distribution 63 shown in FIG.
It protrudes in the positive and negative directions corresponding to the position 3c.

Returning to FIG. 1, the multiplier 49 receives the first-order differential signal J from the first-order differential calculation section 52, and also receives the first-order differential signal J from the first-order differential calculation section 53.
A divider 50 divides the multiplied output signal L from the multiplier 49 by a predetermined constant (2 to the nth power) and sends the divided output signal M to the adder 51. Output to.

The adder 51 inputs and adds the grayscale image signal VD from the A/D converter 47 and the division output signal M from the divider 50, and thereby adds the grayscale image signal VD from the A/D converter 47 to the binarization circuit 43.
generates an output signal N to the

FIG. 8 shows a preprocessing output (output signal N to the binarization circuit 43) obtained by preprocessing the grayscale image signal having the density distribution 63 shown in FIG. 5 by the preprocessing circuit 42 having the above configuration. According to the figure, the edge portion 6 in the density distribution 66
6c has a steeper slope compared to FIG.

The binarization circuit 43 inputs the output signal N of the adder 51, binarizes it using a predetermined binarization threshold, and outputs the binary image signal to the feature measuring section 44. The feature measurement unit 44 measures the area of black pixels in the binary image and performs pattern matching to recognize the shape of the object.

FIG. 2 shows the feature measuring section 4 that performs area measurement.
4, which shows a specific example of area determination section 54 and counter 5.
5 and a register 56. The area determination unit 54
A predetermined window is set for the value image, and it is sequentially determined whether each constituent pixel of the binary image included in the window is a black pixel, and if it is a black pixel, a pulse of level "1" is emitted. Output to counter 55. The counter 55 counts the pulse input from the area determining section 54 to measure the area of the image portion of the object, and the register 56 latches the measured value of the counter 55 and outputs it to the CPU 45.

FIG. 3 shows a specific example of the feature measuring section 44 that performs pattern matching.
7, a match determination circuit 58, a counter 59, and a register 60. A standard pattern is registered in the pattern memory 57, and a match determination circuit 58 inputs the binary image signal of the object and the binary image signal of the standard pattern read out from the pattern memory 57, and compares them pixel by pixel. , when the respective pixel data do not match, outputs a detection pulse of level "1" to the counter 59. The counter 59 counts the detection pulses from the coincidence determination circuit 58 to measure the number of mismatched pixels, and the register 60 latches the measured value of the counter 59 and outputs it to the CPU 45.

Returning to FIG. 1, the CPU 45 executes various calculations related to image processing and controls the operations of each of the above components, and the synchronization signal generator 46 generates synchronization signals. It is applied to each circuit part of the camera 41 and the image processing device 40.

FIG. 9 shows a processing procedure of the image processing device 40 executed in synchronization with the synchronization signal. First, in step 1 (indicated by "ST1" in the diagram) of the same figure, the image processing device 4
0 inputs the grayscale image signal VDi from the camera 41, and in the next step 2, the preprocessing circuit 42 performs the above-described preprocessing on the grayscale image signal VDi. In the next step 3, the binarization circuit 3 binarizes the output signal of the preprocessing circuit 42 to generate a binary image, and in step 4, the feature extraction unit 44 measures the feature amount and performs pattern matching on the binary image. Implement.

FIG. 10 shows a grayscale image 67 whose edges have been emphasized through preprocessing by the preprocessing circuit 42.
In the figure, 68 indicates an image portion of characters, and 69 indicates an image portion of a background. In the pre-processing circuit 42, the product obtained by multiplying the absolute value of the first-order differential of the gray-scale image signal VD calculated by the differential calculation unit 48 by the second-order differential value is used as the gray-scale image signal V.
In order to add the noise component to A desirable grayscale image 67 is obtained in which the edges are not emphasized and the edge portions are emphasized.

FIG. 11 shows the horizontal line 70 of the grayscale image 67.
A brightness distribution 71 along the . In the same figure, 71a, 71b
The part indicated by 2 corresponds to the edge part, and this edge part has a steep slope.

Further, in FIG. 11, 72 indicates the brightness distribution when the imaging conditions change due to illumination deterioration.
The brightness level is lower overall than the brightness distribution 71 under proper illumination.

When each grayscale image is binarized using the binarization threshold value TH shown in FIG. 11, for example, the binarization level along the horizontal line 70 becomes as shown by reference numeral 73 under proper illumination, and when the illumination deteriorates. Then, as shown by the symbol 74,
The widths d of the "0" level are almost the same. Further, the noise portion shown in FIG. 19 does not appear at all in the binarized level of the same figure.

FIG. 12 shows a binary image 75 under proper illumination and a binary image 76 under degraded illumination in comparison. In the figure, the blacked-out part shows a black pixel area representing the character image part 77, and the other parts show the white pixel area representing the background image part 78, respectively. According to the figure, the thickness of the character image portion 77 in the binary image 76 under degraded illumination and the binary image 75 under proper illumination is almost the same, and therefore When a predetermined window 79 is set and the area of the black pixel area within the window 79 is measured to perform shape recognition, etc., the measured area values are almost the same under appropriate lighting and when illumination is degraded, and proper recognition is possible. will be held.

FIG. 13 shows a standard pattern 80 obtained by capturing an image of the letter "A" under proper illumination, and a binary image 81 obtained by capturing an image of a recognition object under degraded illumination. This binary image 81 almost matches the standard pattern 80, and when pattern matching is performed by comparing the two, as shown in the matching image 82 in the same figure, mismatched pixels (shown in black in the figure) (part) will occur, and proper recognition will be made.

[0047]

Effects of the Invention As described above, the present invention provides a preprocessing circuit for a binarization circuit with a differential operation section, a multiplication section, and an addition section, and calculates the absolute value of the first differential and the second differential value of a grayscale image signal. Since the product obtained by multiplying by is added to the grayscale image signal, preprocessing is realized in which the edge portion of the grayscale image is emphasized but the noise component is not emphasized. Therefore, even if illumination fluctuations occur, it is possible to accurately measure feature quantities and perform pattern matching, which is a remarkable effect that achieves the purpose of the invention.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a schematic configuration of an image processing system using an image processing apparatus according to the present invention.

FIG. 2 is a block diagram showing a specific example of a feature measurement unit.

FIG. 3 is a block diagram showing another specific example of the feature measurement unit.

FIG. 4 is an explanatory diagram showing the principle of a differential calculation section.

FIG. 5 is an explanatory diagram showing the density distribution of a grayscale image.

FIG. 6 is an explanatory diagram showing calculation output by a first-order differential calculation section.

FIG. 7 is an explanatory diagram showing calculation output by a second-order differential calculation section.

FIG. 8 is an explanatory diagram showing preprocessing output by a preprocessing circuit.

FIG. 9 is a flowchart showing a processing procedure of the image processing device.

FIG. 10 is an explanatory diagram showing a grayscale image whose edges have been enhanced through preprocessing.

FIG. 11 is an explanatory diagram showing the brightness distribution along a predetermined horizontal line of the grayscale image in FIG. 10;

FIG. 12 is an explanatory diagram showing a comparison between a binary image under proper illumination and a binary image under degraded illumination.

FIG. 13 is an explanatory diagram showing a state in which a binary image is pattern matched with a standard pattern.

FIG. 14 is an explanatory diagram showing a grayscale image input to an image processing device.

FIG. 15 is an explanatory diagram showing the brightness distribution along a predetermined horizontal line of the grayscale image in FIG. 14;

FIG. 16 is an explanatory diagram showing a comparison between a binary image under proper illumination and a binary image under degraded illumination.

FIG. 17 is an explanatory diagram showing the result of pattern matching a binary image with a standard pattern.

FIG. 18 is an explanatory diagram showing a grayscale image with edge emphasis.

FIG. 19 is an explanatory diagram showing the brightness distribution along a predetermined horizontal line of the grayscale image in FIG. 18;

FIG. 20 is an explanatory diagram showing a comparison between a binary image under proper illumination and a binary image under degraded illumination.

FIG. 21 is an explanatory diagram showing a state in which a binary image is pattern matched with a standard pattern.

[Explanation of symbols]

40 Image processing device 42 Pre-processing circuit 43 Binarization circuit 45 CPU 48 Differential operation section 49 Multiplier 51 Adder 52 First-order differential calculation section 53 Second-order differential operation section

Claims (1)

[Claims] 1. A preprocessing circuit that inputs a grayscale image signal obtained by imaging an object and performs preprocessing; and 2 that performs binarization processing on the output signal of the preprocessing circuit to generate a binary image signal. the preprocessing circuit inputs the grayscale image signal, calculates the absolute value of the first derivative and the second derivative, and outputs the first derivative signal and the second derivative signal. a differential calculation section; a multiplication section that receives and multiplies the first-order differential signal and second-order differential signal from the differential calculation section; and adds the multiplication output signal from the multiplication section to the grayscale image signal to generate the binary signal. An image processing device comprising: an addition section that generates an output signal to a conversion circuit.