TW201030675A - Peeling skeletonization image processing method - Google Patents

Abstract

A peeling skeletonization image processing method includes steps stated below. An original image of a character is acquired, wherein the original image has a plurality of image pixels arranged in the form of a matrix. The data of each image pixel is digitalized for converting the original image into a binary image. The data of the image pixels of the binary image is sequentially read according to four directions of the matrix for finding a plurality of marginal image pixels. The data of the surrounding image pixels of the marginal image pixels are used for determining if converting the marginal image pixels into white image pixels. The method continues until the width of each line forming the character becomes the width of a single pixel, so as to convert the original image of the character into a skeletonization image.

Description

[Technical Field] The present invention relates to an image processing method, and more particularly to an image processing method for stripping a skeleton type. [Prior Art] In many engineering systems, image recognition technology is required as a test or judgment of different steps and blocks in the system. For example, in the semiconductor system 1, after each process is completed, it is necessary to use an image recognition system to inspect the workpiece of the gas component to determine whether the process yield is met, for example, to check the line width of the photoresist. Line spacing and so on. In addition, taking the keyboard as an example, all the text symbols on the keyboard, such as English letters, phonetic symbols, etc., are printed on the keys after all the keys of the keyboard are assembled. After the printing is completed, it is necessary to check these. Whether the text symbol is printed correctly can be sold at the factory. With the development of image recognition technology, the accuracy and processing speed = requirements are increasing. Therefore, how to take the image in a more effective and fast way is a problem worth exploring. SUMMARY OF THE INVENTION The present invention relates to an image processing method for stripping a bamboo skeleton type, which is to strip pixels in a symmetrical manner layer by layer, and to frame the image as a line of the width of the pixel, thereby greatly simplifying the original The pixel of the image allows the image of the shirt to be stored in a more streamlined and efficient manner. According to an aspect of the present invention, a method for image processing of a thinning skeleton is proposed, which comprises the steps of: capturing one original image of one character, wherein: 3 201030675 iw)worn original image has multiple pixels, and these Pixels are arranged in an array; binarizing the data values of the pixels to convert the original image to a binarized image; sequentially reading the ^ ', ' in the four directions of the array The data value of the pixel of the image is obtained to find the three edge pixels of the character in these directions, and according to the data values of the surrounding pixels of the edge pixels, whether to convert the edge pixels to blank, = and, repeat the previous step until the line width in the image is reduced to the width of one pixel, thereby converting the original image of the character into a one-inch image. The above content of the present invention can be more clearly understood. The following is a detailed description of the preferred embodiment of the present invention, and is described in detail below with reference to the accompanying drawings: [Embodiment] Please refer to FIG. 1 for the toughness, which is shown in accordance with A flow chart of an image processing method for a shadowy moon frame according to a preferred embodiment of the present invention. The sculpt-like skeletonization = image processing method includes steps 810 to S4 〇 e first, as in step si 〇 image _ take one of the original images of a character, wherein the original image has multiple steps, points, ! The four pixels are arranged in an array. Then, if the step is not, binarize the data values of these pixels to make the original image into a matrix, and to binarize the image. Then, as shown in step S30, the data values of the pixels of the binarized image are sequentially read along the direction of finding== four directions, and the plurality of edge pixels in the direction of the prime character are in the direction, and according to the edge images. The data value of the point m point of the week is to convert the edge pixel to create a blank pixel point. Then, as shown in step S4G, repeating the width of the line in the previous touch-to-image is reduced to the width of one pixel to convert the original image of the character 201030675 X TT 1 into a skeletonized image. Referring to FIG. 2, the above step S30 further includes a plurality of steps S31 to S35. As shown in step S31, the edge pixel of the character in the first direction is found along the first direction of the array, and the first direction is determined according to the data value of the pixel point around the edge pixel in the first direction. The edge pixel is converted into a blank pixel to convert the binary image into a first image. Then, as shown in step S32, the first image is read along the second direction of the array to find the edge pixel of the character in the second direction, and the pixel around the pixel according to the edge of the second direction The data value of the point is used to determine whether the edge pixel in the second direction is converted into a blank pixel to convert the first image into a second image. Then, as shown in step S33, the second image is read along the third direction of the array to find the edge pixel of the character in the third direction, and the pixel points around the edge of the pixel according to the third direction The data value is used to determine whether the edge pixel in the third direction is converted into a blank pixel to convert the second image into a third image. Then, as shown in step S34, the third image is read along the fourth direction of the array to find the edge pixel of the character in the fourth direction, and the pixel around the pixel according to the edge of the fourth direction The data value of the point is used to determine whether the edge pixel in the fourth direction is converted into a blank pixel to convert the third image into a fourth image. In a preferred embodiment, the second direction of the array is opposite the first direction, and the third direction is perpendicular to the first direction, and the fourth direction is opposite the third direction. After the four directions are converted, as shown in step S35, it is determined whether the line width in the image is the width of one pixel, and if so, 5 201030675 indicates that the image has been completed for 4 winds ^

The character fetched in step S1Q in this embodiment is, for example, the character η whose original image is, for example, an 8-value value, between 0 and 255 as shown in Fig. 4A. - an 8-bit image outside, and its data value (between grayscales 5. Binarized image after binarization 1 〇〇 binarized image 100 pixel Pi, j (in which Step S20+, when binarizing, first assigning a plant and then comparing the data value of the pixel with the value of (10). If the data value of a pixel is greater than or equal to the threshold, the pixel is represented. To be 1 or to represent the pixel as a white background, the data value can be set to 〇; when the material of the right pixel is at the reading value, it means the pixel is a dark point or represents the character. The position of some of the lines, the data value * is also shown in the array below the 4A picture, the data value corresponding to the pixels of the binary image is 】 or 〇, the data value of the location is !, the rest Then, the line is as shown in step S401 of Fig. 3, and the data value of the edge pixel of the currently processed image is read. As shown in Fig. 4A®, the word TCH is found along the first direction D1 of the array. The edge pixel of one direction D1 is, for example, 201030675 Μ. Τ 1 «/wvrx Λ- Λ. Pixels Ρ2,12,Ρ3,12 ?4,12 to?3,12,?2,4,?3,4 to?7,4,Ρΐ1,4, Ρΐ2,4 to Ρΐ6,4. When converting, these edge pixels are read and processed one by one. Next, as shown in step S402, it is determined whether it is a processable pixel point, wherein each time an edge pixel point is to be converted, the edge pixel to be converted is located at a center position of a 3x3 matrix, and according to the bit position. In the data values of the surrounding pixels of the other eight positions of the 3x3 matrix, to determine whether to convert the edge pixels to be converted into blank pixels. Taking the pixel points Ρ2, ΐ2 as an example, the 3x3 matrix centered on it The data ❿ [0 0 0, the value can be expressed as 1 1 0, of which the data of the first column and the third row

The value of I1 1 〇J is 〇, and it can be judged that the pixel point P2, 12 is not located at the position of a single pixel point, so it is a processing edge pixel point. Then, as shown in step S403, it is judged whether or not the processed pixel point is a corner point. When the processed pixel point is a corner point, if the pixel point is converted into a blank pixel point, the character line is caused to have a corner shape, and therefore, if the pixel point is a corner point, preferably Keep this pixel. In this step, when the data of the surrounding pixel points of the edge pixel to be converted conforms to a corner point data, the edge pixel to be converted is retained. For example, the corner point data includes the following conditions: '0 0 0, '0 1 0, '0 10, 0 0 0, 0 1 > 0 1 1 0 1 0 , 0 1 oy , 0 0 oy , 0 0 oy , 0 1 0, the blank position in the above matrix is the data value corresponding to the pixel to be converted. If the data of the surrounding pixels meets any of the above conditions, it indicates that the pixel is a corner point without conversion, and returns to step S401 to read 7 201030675 w^>u\J〇rrv to remove an edge pixel p2, 12 and its SI pixel points. Since the pixel is taken, this proceeds to the next step. The body value does not meet any of the above conditions, because the recipient, such as step s4 is the connection point. When the pixel :::: even::: The pixel to be converted is the problem if the pixel is line = discontinuous. Therefore, when you want to convert m2, watch him. In this step, although the pixel of the edge pixel to be converted (4) is used, when the data is clicked, the 保 铍 quot quot 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓 枓For example, even ^ 1 °1 (〇 〇 (0 1 〇x 0 4 , 〇 1 〇 x 1 X, 0 0 X x X Ο X χ χ x χ

J

J

1〇1 V χ ο χ, X 1 χ 0 χ. 'χ Ο 1 χ XXX XXX Ο 1° 1 Ο χ X χ Ο χ Ο χ \ '0 10) \ 1 1 J 〇0, 1 \ 0 X, \ 1 X ) , x 0 X> XXX ο X Ox X The blank position on the material value corresponds to 1 2 1 ^ h represents the data value of the unpaired data 1 豕 (4) $ == pixel point is - connection point without continuation ;; = any, return to step S4G1, read the next edge, change (or remove), ϋ Ρ 3,12. The edge pixel is, for example, imaged at the pixel point Pm and its surroundings. The gastric material value indicating the pixel point ρ(3) is also connected to the point. Therefore, as in step S4G5, ^ is not a corner point 'T 'Remove the pixel point 201030675 1 tv i λ is the pixel value ~ 2 its data value! Convert to 〇 to change to a white background, and store an additional 1 hour image instead of the target H Ρ 2, 12 image for the conversion of the bottom-edge pixel. In the original figure of the fourth figure, the pixel point P2, the position of the u is conjugated ★ ridicule as an example 'square pixel... corresponding data value white background, the pattern is followed by 'as indicated by the step leak, to determine whether to end the Conversion of prime points. Since there are still other edges in the first direction (1): like ~, 1 > 4, 12 to P16, 12, etc. are not processed, so returning to step S401, the data value of the prime point-edge pixel points is, for example, pixel reading. Taking off and further determining whether to repeat the above-mentioned steps 8402 to 8406, the edge value of all the edge pixels in the first direction D1 is processed in the image 200 and its data value as shown in the fourth block, and the other direction is read. Edge pixels: No. Then, as in step (10), the edge pixel in the second direction is processed, wherein the first image of the fourth image in FIG. 4B has a second direction d=, and the edge pixel in the second direction D2 is, for example, a pixel. Point p2, 2, p = 15'2, P16, 2, etc. Similarly, the escape stepped screws S4〇1 to S4〇6 are repeated to process the edge pixels of the first image in the second direction D2, and the second image 300 stored after the processing is as shown in Fig. 4C. Then is to process the edge pixels in the third direction D3, such as pixels

Pi6'3, P1G, 4, P1(), 5 to p1G, 1G and p16 u. The pixel value to be processed 3 '0 10, or pi6, u and its surrounding pixel point data value can be expressed as 0 0 0 This data value can be determined that the edge pixel points p16 3 and P1M1 are all in single 9 201030675 x yy ^ woi i~v The position of a pixel width, so no need to refine it. Direct processing of other pixels, such as pixel point P1M. The above steps S401 to S406 are also repeated to process the edge pixels in the third direction D3, and the third image 400 stored after the processing is as shown in FIG. 4D, wherein the pixel points are? 10, 4 to P10, 10 have become white background, the corresponding data value is 0 °, and finally the edge pixels P2, 3, PM, P8, 5 to Ρ8, ιο and Ρ2 for the fourth direction D4 , ιι to do the conversion. Since it is the same as the foregoing steps, the description will not be repeated, and the fourth image 500 stored after the conversion and its corresponding data value are as shown in FIG. 4E. When the edge pixels in all four directions are converted, as shown in step S35 of Fig. 2, it is judged whether the line width in the image is the width of one pixel. Since all the line widths of the fourth image 500 of Fig. 4E are the width of a single pixel point, it indicates that the skeletonization process of the image has been completed. If not, then return to step S31 to continue the conversion of the next layer of pixels until the image is skeletonized into lines having a single pixel dot width. The image processing method for stripping the skeleton skeleton disclosed in the above embodiment of the present invention first converts the original image into a binarized image, and then strips the pixels of the image layer by layer in a symmetrical manner, for example, from two In the opposite direction, the edge pixels of the image line are stripped, and the edge pixels are stripped from the other two opposite directions, and the steps are repeated, and the image is skeletonized into a line of pixel width. All graphical structural information of the binarized image, such as the position, direction and length of the line, can be preserved at the same time. In this way, the pixels of the original image can be greatly simplified, so that the image can be stored in a relatively simple and efficient manner. As a result of 201030675

丄” ywi»!· r"V Skeletonization and line extraction can greatly reduce the amount of data storage, and can automatically establish the adjacent position and relative relationship of points and lines, and can be further applied to automatic inspection systems, image recognition systems, etc. In order to improve the efficiency of the image recognition, the present invention has been disclosed in the above preferred embodiments, but it is not intended to limit the present invention. Those skilled in the art to which the present invention pertains, without departing from the invention. In the spirit and scope, the various modifications and refinements may be made. Therefore, the scope of protection of the present invention is defined by the scope of the application and the scope of the application. [Simplified description of the drawings] Figure 1 shows A flow chart of a method for processing an image of a thin-skinned skeleton according to a preferred embodiment of the present invention. FIG. 2 is a schematic view showing a step S30 of FIG. 1 further including several steps. FIG. 3 is a diagram showing the steps of FIG. The conversion of edge pixels in different directions further includes a schematic diagram of multiple steps. 10 Figures 4A to 4E, which illustrate a stepwise conversion of a binarized image and its corresponding data values. The main element REFERENCE NUMERALS 100: binarized image 200: first image 300: Second image 400: third image 500: fourth image 11

Claims (1)

201030675 1 VV ^V/UOl ΓΎ ' VII, the scope of application for patents: 1. An image processing method for stripping the skeleton-shaped skeleton, comprising: a. capturing one original image of one character, wherein the original image has a plurality of original images Pixels, and the pixels are arranged in an array; b. binarizing the data values of the pixels to convert the original image into a binarized image; c. sequentially following the array Reading the data values of the pixels of the binarized image in four directions to find a plurality of edge pixels of the character in the directions, and according to the data values of the surrounding pixels of the edge pixels Determining whether to convert each of the edge pixels to a blank pixel; and d. repeating step c until the line width in the image is reduced to a pixel width to convert the original image of the character into a skeleton image . 2. The image processing method of claim 1, wherein the data values of the pixels of the binarized image are 1 or 0. 3. The image processing method of claim 1, wherein the step c comprises: ◎ cl. finding a pixel pixel of the character in the first direction along a first direction, and according to the The edge pixel of the first direction points the data value of the surrounding pixel point, and determines whether the edge pixel of the first direction is converted into a blank pixel point, and the binary image is converted and stored as a first image; c2. Reading the first image along a second direction to find an edge pixel of the character in the second direction, and determining whether the data value of the pixel point around the edge pixel of the second direction is The edge pixel of the second direction is converted into a blank pixel to convert the first image and store 12 201030675 as a second image; c3. reading the second image along a third direction to find the The character is in the edge pixel of the third direction, and according to the data value of the pixel point around the edge pixel of the third direction, determining whether to convert the edge pixel of the third direction into a blank pixel to Second shadow And converting to a third image; and c4. reading the third image along a fourth direction to find an edge pixel of the character in the fourth direction, and according to the edge image of the fourth direction • The data value of the pixels around the pixel is determined to determine whether the edge pixel in the fourth direction is converted into a blank pixel to convert the third image into a fourth image. 4. The image processing method of claim 3, wherein the second direction is opposite to the first direction, the third direction is perpendicular to the first direction, and the fourth direction is The three directions are opposite. 5. The image processing method of claim 3, wherein each time an edge pixel is converted, a temporary image is stored, and the following ® - edge pixel point conversion is performed. 6. The image processing method of claim 3, wherein each time an edge pixel is to be converted, the edge pixel to be converted is located at a center of a 3x3 matrix, and The data values of the surrounding pixels of the other eight positions of the 3x3 matrix are used to determine whether to convert the edge pixels to be converted into blank pixels. 7. The image processing method of claim 3, wherein when the data of the surrounding pixel points of the edge pixel to be converted conforms to a connection point data, the edge pixel to be converted is retained. The image processing method of claim 3, wherein when the data of the surrounding pixel points of the edge pixel to be converted conforms to a corner point data, the method is retained. The edge pixel to be converted.