KR20120074702A - Diagonal scan method for fast edge detection - Google Patents

Abstract

PURPOSE: A diagonal scan method for outline detection is provided to shorten time for analyzing large image data by replacing an inspection with eyes with an image processing technique. CONSTITUTION: Image data of an image are inputted wherein the image is incident through a lens of a camera. A mask is selected. The mask is used when the inputted image data are scanned. Image data are diagonally scanned using the selected mask(S2). When image data are arranged in M x N matrix on a plane, the image data are diagonally scanned from an upper left side to a lower right side.

Description

DIAGONAL SCAN METHOD FOR FAST EDGE DETECTION

The present invention relates to an edge detection technique in the field of image processing, and more particularly, to a diagonal scanning method for edge detection that can minimize the effects of lens distortion of a camera collecting an image.

In the production sites of the display industry (LCD, OLED, PDP, etc.), there is a process of visually inspecting the produced display panels by the operator. As the LCD technology advances, the size of the LCD to be inspected increases and the resolution becomes higher and higher. Visual inspection by the worker has reached a difficult situation. In other words, if the user keeps watching the screen of the LCD which is lit in the process of detecting the error of the produced LCD panel, the worker's eyes are tired and productivity is lowered, and the reliability of error detection is inferior due to individual differences (skills). There is a problem. Therefore, in order to increase the reliability of error detection using standardized values and increase the productivity of LCD equipment in a production line, an image processing inspector using a camera is required.

In order to collect an image using an image processing inspector, a camera is essential. In the camera, distortion occurs in the collected image due to characteristics of an image sensor (CCD sensor, etc.) and a lens. For example, based on the Full HD LCD model, approximately 2,073,600 (1920x1080) pixels should be detected, and the size of approximately 2 million pixels based on 30 inch LCD is about 363㎛ × 363㎛ and one pixel. Has three Sub Pixels. Therefore, at least the CCD cell of the camera and the sub-pixels of the LCD must have about 6 million pixels for 1: 1 matching, and if the calculation is performed with 4 divisions, which is the minimum pixel division, to prevent 1: 1 matching, The required cell of the CCD requires a CCD with at least 6 million x 4 divisions = 24 million pixels. In this case, a large amount of data of about 24M Bytes (2,073,600x12) should be analyzed even if data using at least 8 bit gray level is assumed.

Meanwhile, an image warping technique is used as an image processing technique used to change the position and shape according to the edge distortion of the camera lens (Lens). However, if the position and shape transformation is given by using image warping technique, the original image is damaged due to the real-time operation that is processed during image warping, and it takes a lot of computation time. Programming also involves complex problems because it uses the graphics card's GPU for fast real-world computation. Due to these problems, there is still a quality inspection section using the operator's eyes in the LCD line consisting of the most advanced automatic production facilities.

The present invention has been proposed to solve the above problems, and an object of the present invention is to improve the analysis time of a large amount of image data to replace the visual inspection by the operator in the field of display inspection with image processing technology It is to provide a diagonal scanning method for detecting the contour.

In order to achieve the above object, the method of the present invention comprises the steps of: inputting image data of an image incident through the lens of the camera; Selecting a mask to be applied when scanning the input image data; And scanning the image data diagonally using the selected mask.

이거나

인 것을 특징으로 하고,The mask is

Or

Characterized by

The step of scanning in the diagonal direction is scanning diagonally from the upper right side to the lower left side while sequentially proceeding from the upper left side to the lower right side when the image data is arranged in the M × N matrix form on the plane.

According to the present invention, the contour is detected while minimizing the distortion by the camera lens, and the feature points are extracted to obtain the characteristics of the object (LCD pixel) and labeling, and the error of the LCD pixel compared to the previously stored reference image. Can be extracted. Therefore, the present invention makes it possible to replace visual inspection by the operator in the display field with image processing technology.

According to the present invention, the time required for image processing can be greatly reduced by using a mask during diagonal scan.

1 is an example of a CCD used in the camera to which the present invention is applied.
2 is an example of moiré phenomenon occurring on a CCD screen;
3 is a view for explaining the effect of the lens in the image acquisition through the CCD camera,
4 is a view for explaining a lens distortion and camera position error,
5 is an example of a diagonal scan according to the present invention;
6 is a schematic diagram of an image processing inspector according to the present invention;
7 is a flowchart illustrating an overall procedure for processing an image according to the present invention;
8 is an example of verifying the contour direction of a mask used in a diagonal scan according to the present invention.

The technical problems achieved by the present invention and the practice of the present invention will be more clearly understood by the preferred embodiments of the present invention described below. The following examples are merely illustrative of the present invention and are not intended to limit the scope of the present invention.

1 is an example of an image sensor (CCD) used in a camera to which the present invention is applied, and FIG. 2 is an example of a moiré phenomenon occurring on a CCD screen.

First, the present invention should consider the characteristics of the CCD, the characteristics of the camera lens (Camera Lens), and the characteristics of the TFT-LCD well, so the characteristics of the large-capacity CCD and lens of the camera to obtain the image of the TFT-LCD and TFT-LCD Need to understand.

In the CCD image sensor used in the camera, as shown in FIG. 1, electrons are generated according to the amount of photon of light in the CCD, which is a collection of photodiodes, thereby forming image information forming a screen. Due to this characteristic, the moiré phenomenon as shown in FIG. 2 occurs when the LCD screen is inspected. That is, when the lattice shape of the LCD pixel and the photodiode of the CCD overlap one to one, the charge of the photon is biased to one side, and an interference fringe generated by the overlapping of wave patterns occurs. This interference fringe is a virtual image generated in the process of making an electrical signal. To solve this problem, magnification of the LCD pixel is essential. Increasing the magnification narrows the area that can be inspected, while increasing the number of objects to be inspected, which inevitably leads to a large capacity of the image data.

In addition, the magnification must be enlarged to analyze LCD image data, and if the color filter of CCD is used to receive the LCD's R, G, and B sub-pixels, R, G, B 24bit It needs to be calculated as a row file. Then, the size of the image data file is 12 times {4 (4 pixels of CCD) x because 4 pixels (R, G, B, G) of the CCD operate as one pixel. 3 (Sub-Pixel of the LCD)} as large as the problem occurs. Therefore, in the embodiment of the present invention, R, G, and B are examined only with pure gray levels without using a color filter.

3 is a diagram for explaining the effect of the lens in the image acquisition by the CCD camera.

Looking at the characteristics of the lens of the camera, when the magnification of the image to examine a large area to use a wide aperture lens (Lens). As shown in FIG. 3, the position and image of the center portion of the image passing through the center of the lens and the edge portion of the image passing through the edge of the lens are distorted due to the characteristics of the lens. In order to solve this problem, Image warping method of image processing should be used, but this method takes a long time to process large data due to real operation, and damages the original image. do. Accordingly, the present invention uses a diagonal scan method. In addition, even though the object of FIG. 3 has the same luminance as a whole, the luminance of the image formed on the film side is changed according to the amount of light moved through the lens. The difference between the center brightness and the edge brightness of LCD is about 100 at Gray 0 ~ 255. However, the slope value measured when detecting the contour is constant regardless of the position.

On the other hand, unlike the general object seen reflected by the sunlight, in the case of TFT LCD, LED or CCFL light source is seen through the liquid crystal. This liquid crystal usually responds at an operating speed of 60 Hz. However, the problem is that the liquid crystal deterioration phenomenon that the liquid crystal does not work anymore because the TFT LCD keeps applying the same voltage by direct current and the liquid crystal keeps standing in the same direction and the liquid crystal panel is pushed to one side along the polarity. Occurs. Therefore, it solves the 'liquid crystal deterioration' problem by continuously changing + and-when applying voltage on the LCD panel. For example, TFT-LCD is designed to continuously change the voltage so that the liquid crystal can be tilted by 45 degrees and 45 degrees to the left and 45 degrees to the right in the next frame. At this time, the same polarity should be applied. Because of the parasitic capacitance with the cap, it is not easy to adjust the angle from the left and right, which makes the screen flicker. As such, the intensity of light emitted from the LCD screen is not constant but changes periodically with time, so that a flicker of light is felt by the user.

Because of this flicker, the camera does not emit light at the same time in all pixels of the LCD, so the exposure time of the camera must be set sufficiently. For example, a TFT-LCD driving at 60 Hz requires an exposure of at least 33 ms (16.6 x 2 = 33.3 ms) since the camera exposure time is at least 33 ms (60 Hz = 16.6 ms but tilted left and right).

However, if the exposure time increases due to the characteristics of the LCD pixel composed of R, G, and B sub-pixels, green light spreads out from bright light due to the brightness, which may result in difficult to analyze image data. . Therefore, adequate testing should ensure proper camera exposure time.

4 is a diagram illustrating a lens distortion and a camera position error, and FIG. 5 illustrates a diagonal scan method according to the present invention.

If the position of the LCD pixel is concave as shown in FIG. 4A due to the distortion of the camera lens or the angle of the image is slightly changed due to an error in positioning, the position of the image to be inspected is changed as shown in FIG. Scanning with will make it impossible to inspect the pixels in order. Accordingly, in the present invention, as shown in FIGS. 5A and 5B, image data may be scanned in a diagonal direction, and pixels may be sequentially searched even if the position of the image changes according to the distortion of the lens.

5A and 5B, when image data is arranged in an M × N matrix form on the x and y axis planes, scanning is performed in a diagonal direction while sequentially proceeding from the upper left side to the lower right side. That is, as shown in FIG. 5A, the scan is performed in the diagonal direction from the upper right side to the lower left side as shown by the arrow in the order of 1,2,3,4,5,6,7,8,9.

As described above, the image between the center of the LCD and the image of the corner is substantially the same, but the camera using the CCD of 10M Pixel or more has a large aperture of the lens, which causes distortion in the size and shape of the image when the image is formed on the CCD. The brightness of the, phase is also different. Correcting the distorted image affects the position of the pixel, and correcting the luminance causes difficulty in detecting an error of the pixel, in which the actual luminance is a problem. In addition, when the noise is summarized, since the original data is modified, the reliability of each pixel information is inferior.

Therefore, in the present invention, the original data is detected from an object (ie, the pixel of the LCD) to be searched for, and then the characteristics (position, size, average luminance, maximum luminance) of each object are well combined to mask the luminance ( Masking method is used.

The algorithm for detecting contours during diagonal scans is performed in the diagonal direction so that a method such as Equation 1 or Equation 2 is used to reduce the time by increasing the speed at the time of scanning.

When the conventional first derivative is used, more than 14 calculations are required in both the X and Y directions, and even when the second derivative is used, a total of nine calculations are required. If the mask of Equation 2 is used, it can be processed by two or three operations, thereby increasing the efficiency of the calculation.

Also, since it is a part that finds the target object, there is no need to apply the existing contour algorithm, and it scans the object downward from left to right and top to bottom. For the purpose of finding, we propose a mask like Equation 1 or Equation 2.

6 is a schematic diagram of an image processing inspector to which the present invention is applied, and FIG. 7 is a flowchart illustrating an entire process of processing an image according to the present invention, and FIG. 8 is a contour direction of a mask used in a diagonal scan according to the present invention. This is an example of verification.

As illustrated in FIG. 6, the image processing inspector to which the present invention is applied is diagonally formed by dividing the image data of an image input from the camera 20 and a camera 20 for capturing image data by capturing the LCD 10 as an inspection object. Diagonal scan module 30 for scanning, contour detection module 40 for detecting the contour by processing the data obtained by the diagonal scan, and feature point extraction module 50 for extracting the size of the detected object (pixel), etc. And a determination module 60 for determining defects of the inspection target by using the extracted feature points. The diagonal scan module 30 includes an image input unit 32 for receiving image data, a mask selecting unit 34 for selecting a mask to be applied during scanning, and a scanning unit 36 for scanning image data in a diagonal direction with the selected mask. )

Referring to FIG. 7, image data is acquired by a camera and then scanned until a contour is extracted by a diagonal scan using a mask (S1 and S2). When scanning, use the diagonal masking operator proposed in accordance with the present invention. When the contour is extracted, the maximum value is obtained in the direction of the contour, and the contour is completed by connecting the points where the slope rapidly changes around the maximum value (S3, S4). When the outline is completed, the size and shape of the object are reviewed, and when all objects are obtained, the point number is found by checking the number of pixels (S5). S6).

FIG. 8 illustrates a form in which various types of objects are scanned when detected by using a mask method proposed according to the present invention. The Mask operator can be operated in the same direction by scanning while moving from top to bottom and right to left.

Referring to Figure 8, it can be seen that the operation is made in the direction of the rectangle, circle, diagonal 1, diagonal 2.

The present invention has been described above with reference to one embodiment shown in the drawings, but those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom.

10: inspection target 20: camera
30: diagonal scan module 40: contour detection module
50: feature extraction module 60: judgment module

Claims (4)

Inputting image data of an image incident through a lens of a camera;
Selecting a mask to be applied when scanning the input image data; And
And scanning the image data in a diagonal direction using the selected mask. The method of claim 1, wherein the mask is

Diagonal scanning method for contour detection, characterized in that. The method of claim 1, wherein the mask is

Diagonal scanning method for contour detection, characterized in that. The step of scanning in the diagonal direction of claim 1
A diagonal scanning method for contour detection, characterized by scanning diagonally from the upper right side to the lower left side while sequentially proceeding from the upper left side to the lower right side when the image data is arranged in a M × N matrix form on a plane.

Images