JPH09259281A - Image information processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a processor for more accurately detecting and processing irregular components inside image information at a higher speed. SOLUTION: Inputted digital image information is stored in an input image memory 31 and the irregular component is separated in an irregular component calculation means 32 and stored in an irregular component image memory 33. Then, binarization is performed by an optional threshold value, a labeling processing is performed, a three-dimensional irregularity value for which a density value is considered is calculated in a three-dimensional irregularity value calculation means for the component of the same label and stored in a three-dimensional irregularity value memory and normal/defective conditions are judged and correction is performed by the three-dimensional irregularity value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus used for an image quality inspection apparatus such as a CCD tester or an LCD (liquid crystal) tester for inspecting an object using an image, or an image recognition apparatus for determining an object by an image. In particular, the present invention relates to an image information processing apparatus that detects and processes a so-called "unevenness component" in image information.

[0002]

2. Description of the Related Art Conventionally, there are many image processing apparatuses that inspect an object on a CCD or LCD using an image, recognize, judge, and diagnose an object of an object.
Many of these digital images handle images digitally due to the spread of computers. A digital image is captured by an image capturing camera such as a CCD, and the pixel density is A /
D-conversion (analog / digital conversion), for example, creates a grayscale image in which the density is digitized from 0 to 100, stores it in a memory, performs necessary processing, stores it in memory, determines whether it is good or bad, and transmits by communication. Finally, it is displayed on the display device. The number of pixels may be arbitrarily determined, but in the VGA standard, the height and width are 640 × 480 = 307,200 points,
800 × 600 points for SGA standard, 10 for GA standard
This is a screen of 24 × 768 points. In the case of a color image, three colors R (red) G (green) B (blue) are handled, so the number of data handled is tripled.

By the way, one of the troublesome things in handling a digital image is so-called "unevenness". The "unevenness" is a luminance change having a certain area on the screen, and refers to a portion where the amount of the luminance change is several% of the density value, which is also a kind of noise. The unevenness may be caused by a specific cause of the CCD element or the LCD element, or may be generated by a light source, an amplifier or the like, or may be induced by these circuits and superposed on the image data at the time of transmission. If the unevenness occurs, the quality of image information is deteriorated, and erroneous determination or erroneous diagnosis may occur. The present invention promptly and accurately detects the nonuniformity component and makes a pass / fail judgment.

First, a general television (TV) screen or LC
A high-quality image on the D screen will be described with reference to FIG. FIG. 6B shows the luminance density level a for each line in the horizontal direction of the screen. Shading (uniformity of brightness)
That. In normal shading, there is a density difference of about 20% to 30% between the central part and the peripheral part of the screen, but since it gradually changes, it can be seen with almost no effect on the naked eye. The image data actually input by the image information processing apparatus includes the image data shown in FIG.
Since noise n is superposed on this, as shown in, the noise is removed by a two-dimensional filter.

FIG. 7 shows a part of a color mixture inspection device for a color cathode ray tube panel as an example of a conventional image information processing device. The camera 11 is an R (red) camera 11 ₁ and a G (green) camera.
It has a camera for camera 11 ₂ and a camera for B (blue) 11 ₃ and digitizes the densities of the respective colors by the digital image input device 12. The digital image input device 12 has an amplifier, a sample hold circuit, an AD converter, etc., as required.

The digitized digital image data is sent to the image information processing device 13, and first, the input image memory 16 is inputted.
Is stored in the memory. Next, noise is removed by filtering with a two-dimensional filter, for example, the average value calculating means 17, and the processed digital image data is stored in the average value memory 18 as R, G,
It is stored for each B and displayed on the display unit 14 through the output means 19 or printed by the printer 19. Moreover, it outputs to the outside through an output terminal as needed.

The image information processing in FIG. 7 is performed by the average value calculating means 1
Although the average filter method according to No. 7 is used, the median filter method may be used. That is, a method of reducing noise in a digital image is a method of neighborhood processing, which uses a two-dimensional filter and replaces the density value of the pixel of interest with the calculation result of the density values of neighboring pixels. As a noise reduction method in the neighborhood processing, a method using an average filter by arithmetic averaging and a method using a median filter that rearranges N data values in order of magnitude and selects a median value are representative.

Further, the density of a pixel is often binarized and displayed as a memo. That is, assuming that the density of the pixel is 0 to 100, the threshold value (threshold value) is set to an arbitrary value, for example, 50, which is exactly half, and the pixel whose density value is equal to or higher than 50 is 1 as the active image and is black. The pixels of 50 or less are displayed in white as inactive 0. The density value of the pixel is binarized to 0 or 1 by selecting an arbitrary threshold value.

Labeling processing may be performed on the one binarized active image. The labeling process is to label one active image, but if there are adjacent active images, the same label is attached. Therefore, the active image has the specific label number information, and the non-active image that is not labeled does not have the label number information. Then, in the active image having the specific label number, it is possible to easily know the number of pixels in each same label, that is, the area of the active image having the same label.

In the CCD tester and LCD tester, the size of the area of the same label is used to judge the quality of the CCD or LCD which causes a unique unevenness defect. For example, when the entire surface receives light in white, is filtered by the image information processing device 13, and is displayed on the display unit 14, it is displayed in white, which is the same as the light-receiving color when normal. However, if the object of the CCD or LCD is a defective product, uneven defects inherent in the screen occur. The quality of this uneven defect was determined by the size of the area.

[0011]

Since the present invention can be applied to all image information processing apparatuses, it is possible to correct unevenness superposed during transmission. However, for the sake of easy understanding, a CCD having a unique unevenness defect is provided. An image information processing apparatus that tests whether or not the LCD is good will be described. As described above, according to the processing method in the image information processing device in the conventional CCD tester or LCD tester, the input image data is taken in and noise including a nonuniformity component is removed by a filter such as an average filter or a two-dimensional median filter. After the removed image data is obtained, the difference between the removed image data and the input image data is obtained to separate the unevenness component, the density of the pixels of the unevenness component is binarized with an arbitrary threshold value, the binarized data is subjected to labeling processing, and The quality of the specific unevenness defect was judged by the size of the specific labeling area.

Although a certain effect has been achieved, if it is attempted to completely remove noise, the calculation will take too much time,
If the processing is roughened and the processing time is shortened, the processing becomes incomplete and the required performance cannot be satisfied. According to the present invention, the processing time for noise removal is relatively short, and the inherent unevenness defect is detected more accurately than before.

[0013]

The first aspect of the present invention is a nonuniformity component calculating means for detecting a nonuniformity component, which uses a new median calculating means. The median calculation is an operation of replacing the density value of the pixel of interest with the median (median) of the density of pixels in the observation window formed by a plurality of two-dimensional pixels centering on the pixel of interest. According to the median calculation method of the present invention, the one-dimensional median is calculated in the horizontal direction and the vertical direction, and the median calculation is performed at high speed and with the same accuracy as the two-dimensional median calculation.

First, the median is calculated in the horizontal direction of the pixel (hereinafter referred to as "H median") to obtain the H median of all the pixels, and then the median is calculated in the vertical direction of the pixel (hereinafter referred to as "V median"). The V median of the pixel is obtained. Conversely, the H median may be obtained from the V median. Hereinafter, this will be referred to as "HV median", and this calculation means will be referred to as HV median calculation means.

When the H · V median is obtained from the input image data, the H · V median data of that value has only the shading waveform from which the noise component including the unevenness component is removed. When the difference between the H.V median data and the input image data is obtained, the shading waveform is removed and the nonuniformity component is separated to obtain the density value of only the nonuniformity component. This will be referred to as uneven component calculation means. The density value of the nonuniformity component image data is sliced with an arbitrary threshold value to be binarized, and if necessary, a single active image is regarded as noise and removed, and labeling processing is performed and stored in the labeling memory. The quality of the CCD or LCD is determined by the number of pixels of the same label, that is, the area.

A second aspect of the present invention is to judge the quality of an object not by the area of the same label, but by a new means, by considering the density value of the same label, by the volume of the same label, that is, the size of the three-dimensional unevenness value. It is something to do. Therefore, a new three-dimensional unevenness value calculating means is provided. The structure of the invention is as follows. The noise component including the unevenness component is separated from the input image data by the unevenness component calculation means. The unevenness component calculating means preferably has the configuration of the first invention, but is not limited to this, and a conventional configuration may be used. That is, a method using an average value calculating means or a median filter method using a two-dimensional observation window may be used. Any material that can remove the shading and separate the uneven components may be used.

When the nonuniformity component is separated, it is binarized with a threshold value,
A single active pixel is removed and a labeling process is performed. Next, the three-dimensional unevenness value is calculated by considering the area of the same label, the maximum density value and the average density value of the label. 3D unevenness value A, the area of the same label A, B and the average density value, the maximum density value is C, the a k _1, k _2, k ₃ and calculation coefficients, a function such as: is there. Three-dimensional unevenness value = f (k ₁ × A, k ₂ × B, k ₃ × C) Here, the coefficients k ₁ , k ₂ and k ₃ are to be obtained while correlating with the visual results depending on the object.

When the three-dimensional unevenness value is obtained in this way,
Since the value indicates a value corresponding to the energy of the unevenness component rather than the area of the same label, the intensity of the unevenness component can be known more accurately. For example, assuming that the area of a plurality of identical labels is 100, the variation of the density value considering 1 to 10
If so, even if the area is the same, the three-dimensional unevenness values will be different values of 100 to 1,000, and more accurate intensity of the unevenness component can be known. Therefore, the quality is judged based on the magnitude of the three-dimensional unevenness value.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below in detail with reference to Examples.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram of an embodiment of the first invention of the present invention, FIG. 2 is a block diagram of an embodiment of the second invention, FIG. 3 is a processing flow chart, and FIG. A diagram and FIG. 5 are explanatory diagrams of the processing steps. Parts corresponding to those in FIGS. 6 and 7 are designated by the same reference numerals. First, the process of FIG. 5 will be described.

Digital image information as shown in FIG. 5A is input to the image information processing apparatus in one line of the horizontal method.
The unevenness components b, c and d are superimposed on the shading waveform a. The high frequency noise component n is omitted. When this input image information is passed through a two-dimensional filter, the noise component including the unevenness component is removed and only the shading waveform a is obtained as shown in FIG. 5B. Therefore, when the difference between the data in FIG. 5A and the data in FIG. 5B is calculated, FIG.
The uneven components are separated as shown in.

As shown in FIG. 5C, the separated unevenness component is sliced at an arbitrary threshold value and binarized, and only the unevenness component above the threshold value is used as an active image. When this is stored in the memory, the entire screen is processed, and the labeling process is performed, the memory map shows only the uneven component part b, as shown in FIG.
The active image has a constant area of c and d. A pass / fail judgment is performed using this labeled active image.

The first invention will be described with reference to the embodiment shown in FIG. The image information processing apparatus 30 is the digital image input apparatus 12
The digital image information is input from and is stored in the input image memory 31. R, G, B when the input image information is color
Processing such as storage / calculation is performed for each of them, but here, a monochrome system will be described. Further, in FIG. 1, the analog image from the camera 11 is A / D converted by the digital image input device as in the conventional example of FIG. 7, but if the transmitted digital image information is directly input. Good.

The nonuniformity component is separated from the input image data stored in the input image memory 31 by the nonuniformity component calculating means 32. The unevenness component calculating means 32 includes an H · V median calculating means 26, a median data memory 27, and a difference calculating means 2.
It is composed of 8. H / V median calculation means 26
As described above, the H median in the horizontal direction is first obtained for the pixel of interest, and when the H medians of all the pixels are obtained, then the V median in the vertical direction of all the pixels is obtained. Since it is a one-dimensional median calculation, it can be calculated at high speed. The number of pixels for median calculation may be about 5 to 9, and the frequency band may be selected by selecting continuous pixels, every other pixel, or every two pixels. Conversely, this procedure may be performed by first obtaining the V median and then the H median. In short, it is important to perform filtering for obtaining both the median in the horizontal direction and the median in the vertical direction for the target pixel on the entire screen.

When the H · V median is calculated, the H · V
Since the V median data is data only for shading with the uneven component removed, the temporary median memory 27
When the difference between the H.V median data and the input image data is obtained by the difference calculating means 28, only the nonuniformity component is separated and the nonuniformity component image data is obtained. When the uneven component image data is obtained, the H / V median data becomes unnecessary, so that the dedicated median data memory 27 is unnecessary.

The nonuniformity component image data is temporarily stored in the nonuniformity component image memory 33, the data is sliced by the binarization calculation means 34 at an arbitrary threshold value, and pixels having a density value equal to or higher than the threshold value are set as an active image. Pixels are inactive images. This binary image is stored in the binary image memory 35. In this binary image memory map, a single active image is regarded as noise and removed as necessary. You may consider several or less connected active images as noise. After that, the labeling means 36 labels the active image. The same label is attached to the connected active images. The labeled active pixels are stored in the labeling memory 37 with information such as coordinate information and the number of active images in the order of labels.

Based on the labeling data, the processing means 39 including the judging means 40 judges the quality of the object.
The result is displayed on the determination display unit 42, the memory map is displayed on the display unit 14 through the output unit 38,
It is printed by the printer 15. A data bus is connected between these various calculation means and various memories to exchange information. The control unit 42 controls the entire system.

FIG. 2 shows an embodiment of the second invention. The parts common to FIG. 1 are omitted, and only different parts will be described. The unevenness component calculating means 32 may use the H · V median calculating means of the first invention, or may use the conventional average value calculating means or two-dimensional median calculating means. The newly provided one is a three-dimensional unevenness value calculating means 51 and a three-dimensional unevenness value memory 52.
Therefore, the unevenness component image memory 33 requires a dedicated one.

The three-dimensional unevenness value calculating means 51 calculates a three-dimensional unevenness value in which the average density value B and the maximum density value C of the same label are added to the area A of the active pixels having the same label. The function is the following equation as described above. Three-dimensional unevenness value = f (k ₁ × A, k ₂ × B, k ₃ × C) This function differs depending on the object, but as an example of the simplest expression, a sum of linear functions may be used as in the following expression. . Three-dimensional unevenness value = k ₁ × A + k ₂ × B + k ₃ × C

These relationships will be described with reference to FIG. FIG.
(A) is a memory map of a binary image that has undergone labeling processing as 1, 2, 3, 4. In the first invention, the quality is judged by the size of the area of the labeled unevenness component. In the second invention, since the density value of the unevenness component image memory 33 is added to the area of the labeled unevenness component, the volume becomes three-dimensional as shown in FIG. 4B. Therefore,
In the three-dimensional unevenness value, the unevenness component has a more correct energy value and the difference between the labels becomes large, so that the determination is easy and the accuracy is improved.

FIG. 3 is an example of a flow chart of the processing steps of the image information processing apparatus of the present invention. Image information is input 61, stored in the input image memory for each pixel 62, the unevenness component is separated using the H · V median calculating means 63, and the density of the pixel is slid at an arbitrary threshold value to be binarized 64. If necessary, the single active pixel is regarded as noise and removed 65, and the adjacent active images are labeled with the same label 65. When performing three-dimensional unevenness calculation, three-dimensional unevenness calculation is performed 68, and not performed. In the case of 67 of the same label area,
It is an example of a flow chart of performing 69 pass / fail judgment. It is not necessary to limit to this flow.

[0032]

As described above in detail, according to the present invention, the H / V median calculating means 2 is used in the image information processing apparatus.
Since 6 is used, the nonuniformity component can be detected and processed faster and more accurately. Further, since the energy of the nonuniformity component can be more accurately grasped by using the three-dimensional nonuniformity value calculating means 51, it becomes possible to more accurately determine the quality of the CCD or LCD having the inherent nonuniformity defect.

As described above, the present invention is an image information processing apparatus capable of improving image quality and eliminating erroneous determination and erroneous diagnosis by being put to practical use.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of a first invention.

FIG. 2 is a configuration diagram of an embodiment of a second invention.

FIG. 3 is an example of a flowchart of a processing process of the present invention.

FIG. 4 is an explanatory diagram of a nonuniformity component.

FIG. 5 is an explanatory diagram of a processing process.

FIG. 6 is an explanatory diagram of image data of a non-defective object.

FIG. 7 is a configuration diagram of an example of a conventional technique.

[Explanation of symbols]

 11 Camera 12 Digital Image Input Device 13, 30, 50 Image Information Processing Device 14 Display Unit 15 Printer 16, 31 Input Image Memory 17 Average Value Calculation Means 18 Average Value Memory 19, 38 Output Means 20, 42, 53 Control Unit 26 H V median calculating means 27 median data memory 28 difference calculating means 32 uneven component calculating means 33 uneven component image memory 34 binarizing calculating means 35 binary image memory 36 labeling means 37 labeling memory 39 processing means 40 determining means 41 correcting means 43 Judgment display section 51 Three-dimensional unevenness value calculation means 52 Three-dimensional unevenness value memory

Claims (3)

[Claims] 1. An image information processing apparatus (3) for inputting digital image information, detecting a non-uniformity component of the input image information, and performing processing.
0), the input image memory (31) for storing the input digital image information, and H.sub.median and V.sub.median of the target pixel are calculated by using the input image data of the input image memory (31).
An unevenness component calculating means (32) for obtaining unevenness component image data by a V median calculating means (26) and a difference calculating means (28) for obtaining a difference between the H · V median data and the input image data, and the unevenness. Binary calculation means (34) for binarizing the component image data with an arbitrary threshold value to obtain binary image data, and a binary image memory (35) for storing the binary image data.
A labeling means (36) for performing labeling using the binary image data, a labeling memory (37) for storing the labeling data obtained by the labeling means (36), and a determination using the labeling data. An image information processing apparatus comprising: a processing unit (39) for performing correction. 2. An image information processing apparatus (5) for inputting digital image information, detecting a non-uniformity component of the input image information, and processing it.
In (0), an input image memory (31) for storing the input digital image information, and a nonuniformity component calculating means for obtaining nonuniformity component image data by separating nonuniformity components using the input image data of the input image memory (31) ( 32), an unevenness component image memory (33) for storing the unevenness component image data, and a binarization calculation means (34) for binarizing the unevenness component image data with an arbitrary threshold value to obtain binary image data. A binary image memory (35) for storing the binary image data
A labeling means (36) for performing labeling using the binary image data, a labeling memory (37) for storing the labeling data obtained by the labeling means (36), the labeling data and the unevenness component image A three-dimensional unevenness value calculating means (51) for obtaining three-dimensional unevenness value data from the density value of the data, a three-dimensional unevenness value memory (52) for storing the three-dimensional unevenness value data, and the three-dimensional unevenness value data. An image information processing apparatus comprising: a processing unit (39) for making a determination or a correction. 3. The image information processing apparatus according to claim 1,
0) and the image information processing device (50) according to claim 2, which is configured by a computer.