KR20000056676A - A parallel thinning method using weighted value - Google Patents

Abstract

PURPOSE: A parallel thinning method using weight value is provided to cut information content requested for automatic recognition of image pattern to minimum information content, by granting weight value related to the number of neighbor pixel to outer pixel and then deleting the pixel satisfied to specific condition from outer line at the same time. CONSTITUTION: A parallel thinning method using weight value comprises the steps of: first determining a weight value of respective black pixel to image pattern inputted from an image input device; second checking deleting condition corresponding to the weight value and deleting the corresponding pixel if the outer pixel is satisfied to the deleting condition; third determining the weight value of respective black pixel;fourth checking deletion condition corresponding to the weight value and deleting the corresponding pixel if the outer pixel is satisfied to the deleting condition; and fifth granting again the weight value until there is no outer pixel satisfied to the deleting condition fifth and repeating the third and fourth step.

Description

A parallel thinning method using weighted value}

The present invention relates to a method for thinning a digital image that can be widely used in various fields such as character recognition, defect inspection of a circuit board, fingerprint recognition, and the like. The present invention relates to a parallel thinning method using weights that simultaneously removes pixels satisfying a specific condition from a target, thereby reducing the amount of information necessary for automatic recognition of an image pattern to a minimum amount.

In general, in the method of thinning a binary image while deleting outer pixels, the following main evaluations should be considered.

First, you must not lose your original shape and topology.

Second, the outer pixel must remain at the minimum unit width.

Third, find the nearest centerline.

Fourth, it should be less affected by noise.

Fifth, the treatment efficiency should be good.

The five requirements described above must be met as much as possible to achieve improvements in efficiency, accuracy and performance.

On the other hand, the thinning of the conventional digital image pattern can be largely divided into a pixel-based iterative outer pixel removal thinning method and a pixel-based thinning method.

The non-pixel-based method finds the centerline directly without examining each pixel.

However, the non-pixel based method has the disadvantage of low computational complexity while producing noise when the stroke is parallel to the scan line. In particular, it is unsuitable for strokes having a wide variety of shapes and thicknesses.

Meanwhile, the pixel-based iterative outer pixel elimination thinning method is a method of repeatedly removing unnecessary pixels until a skeleton line remains, and the thinning method based on such iterative outer pixel elimination is parallel with the sequential thinning method. It can be divided by thinning method.

In the sequential thinning method, the removal of the n-th order pixel is affected by all the pixel removals performed up to that time, but in the parallel thinning method, the deletion of the n-th order pixel depends only on the result of the (n-1) th iteration. .

As such, all pixels are examined in parallel and independently at each iteration. The disadvantage of the general parallel thinning method is that not only neighboring 8 pixels but also the next neighbor (24 pixels) must be searched to maintain connectivity.

In order to solve this problem, most of the parallel thinning methods go through two or four sub- iterations. The recent single iterative thinning method has been developed by Zhang. This thinning method is a thinning method having good thinning ability, such as maintaining connectivity, thinning to unit pixel width, and noise processing capability.

However, this method has a problem of having to check up to 24-neighbor pixels and also having to recover deleted pixels to maintain connectivity.

An object of the present invention for solving the above problems is to give weights related to the number of adjacent black pixels among the black / white pixels of the input image when thinning the outline of the binary image, and to delete all the pixels Instead of retrieving the deletion conditions, only when the detection condition is defined corresponding to each weight, only the corresponding deletion condition is checked, and the corresponding pixel is deleted only when the deletion condition is satisfied. It is to provide a parallel thinning method using a weight to reduce the amount to the minimum amount.

1 is a flowchart illustrating a parallel thinning process using weights according to the present invention.

2A is a pixel block for an input binary image.

2B is a pixel block representing 8-neighboring pixels of pixel p in a 3x3 template.

2C is a pixel block representing 24-neighbor pixels of pixel p in a 5x5 template.

FIG. 2D is a pixel block weighted to FIG. 2A; FIG.

2E is a template block for describing a 3 × 3 template in accordance with the present invention.

3A is a block of pixels for a binary image with a vertical image two pixels thick;

3B is a pixel block for a binary image with a vertical image three pixels thick;

3C is a pixel block weighted to FIG. 3A;

3D is a pixel block weighted to FIG. 3B.

4 is a diagram illustrating weights according to the present invention and templates and deletion conditions according thereto;

In the parallel thinning method using weights according to the present invention for achieving the above object, a weight indicating the number of black pixels adjacent to each black pixel is given to a binary image pattern input to an image processing apparatus through an image input device. And a second step of deleting the corresponding black pixel if the pre-stored deletion condition is satisfied, and a black pixel satisfying the deletion condition. The weighting is again performed until there is no, and the black pixel deleting step is repeatedly performed, but the black pixel having a weight without any black pixels adjacent to the black pixels is performed without deleting even if the black pixel with the weighting condition is satisfied. If there is no black pixel that satisfies the deletion condition, the thinning process is terminated, and finally the thinned image That made in a third step for outputting a turn-characterized.

The above objects and various advantages of the present invention will become more apparent from the preferred embodiments of the invention described below with reference to the accompanying drawings by those skilled in the art.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a parallel thinning process using weights according to the present invention. As shown in the drawing, a binary image input to an image processing apparatus 001 through an image input apparatus 002 such as a scanner or a camera is illustrated. A weight defined in accordance with the present invention is determined and assigned to the pattern 003 (004), and according to the weight, whether or not to delete each pixel is determined by an inspection template pre-stored in the memory 007 and an erase condition. And the weight determining step 004 and the pixel deleting step 005 are repeatedly performed until the thinning end condition 006 is satisfied (005). 008) is obtained as an output result.

This process is summarized in the following steps.

Step 1. Determine the weight of each black pixel for the input image pattern.

Step 2. Since the outer pixels have respective weights K, an erase condition corresponding to the weight K is searched for and deleted if the erase condition is satisfied. Independent pixels are considered noise pixels)

Step 3. Determine the weight of each black pixel.

Step 4. Since the outer pixels have respective weights K, an erase condition corresponding to the weights K is searched for and deleted if the erase condition is satisfied. Independent pixels thinned in turn are considered to be the result of repetition of thinning. For example, when a full sphere is thinned, it is thinned to one pixel. )

Step 5. Repeat steps 3 and 4 until there are no black pixels that meet the deletion criteria.

Next, a process of parallel thinning using weights according to the present invention will be described in detail.

2A is a binary image pattern input through an image input device. The input binary image pattern is divided into a black pixel represented by 1 and a white pixel represented by 0. FIG. That is, the pixel having a value of 1 corresponds to the outline portion.

Each pixel may be represented by a coordinate form of a (row number, column number) pair by using a row number and a column number.

2b and 2c are blocks for 8-neighbor pixels adjacent to each pixel p and 24-neighbor pixels adjacent to each pixel p, in particular a 3x3 block comprising 8-neighbor pixels in FIG. The boundary area for determining the weight is shown.

In other words, the weight employed in accordance with the present invention represents the number of black pixels located in the eight-neighbors of a black pixel. Therefore, the weight has one of 0-8. However, in the case of an independent black pixel having no black pixels among adjacent pixels, the weight of the pixel should be 0. However, since 0 is already used to represent the white pixel, 9 is used as the weight for representing the independent black pixel.

2D illustrates an image pattern in which the input image pattern illustrated in FIG. 2A is weighted according to the present invention.

In the same figure, in the case of the pixels 2 and 2, as shown in FIG. 2A, the number of black pixels among the 8-neighbor pixels is three because of the weight of the pixel 2, 2, and 3, In the case of the pixels 2 and 6, as shown in FIG. 2A, since the pixel 8 does not include any of the adjacent 8-neighbor pixels, the weight is assigned to 9.

Meanwhile, the black pixels having a weight of 1 to 7, that is, the remaining black pixels except for the pixels 5 and 3 are referred to as outer pixels.

2E is a template block for describing a 3 × 3 template according to the present invention, where W (p) represents a weight of an outer pixel and X represents a weight of a neighboring pixel. Therefore, since W (p) is the weight of the outer pixel, it can be seen that X equal to W (p) should exist in the neighborhood.

First, each processor takes care of one pixel and processes the thinning in parallel, where the parallel means having one pixel after one iteration, that is, several thinning steps for an image. The resulting image can be obtained, meaning that each processor can process each pixel simultaneously.

In addition, an erase condition is required for the parallel thinning method of repeatedly removing the outer pixel based on the pixel. The deletion of any black pixel (outer pixel) is determined by the inspection template and the deletion condition associated with the inspection template. Since the parallel thinning method repeatedly removes the outer pixels, its performance depends on the number and size of test templates.

Since the test template formed in accordance with the present invention uses weights, its size and number can also be reduced, and FIG. 4 shows details of the type and deletion conditions of the test template corresponding to the weight used in accordance with the present invention. This data is stored in the memory 007 described above.

That is, the inspection template according to the present invention has a 3 × 3 block size, and if there is an inspection template whose shape matches among the inspection templates for removing the outer pixel while the outer pixel p has a weight w (p), the corresponding delete condition is applied. In this case, that is, if the test template for the weight and the deletion condition are satisfied at the same time, the pixel p can be deleted.

As shown in FIG. 4, the inspection template form and the deletion condition for each weight according to the present invention are summarized as follows.

If the weight is 1, there are eight test templates located in one adjacent black pixel east, west, south, north, northeast, northwest, southeast, or southwest direction, and the delete condition is that the weight of the adjacent black pixel is greater than three. Delete it.

If the weight is 2, two adjacent black pixels are northeast and east, north and northeast, northwest and north, northwest and west, west and southwest, southwest and south, south and southeast, southeast and east, east and north, east and east It has 12 test templates located in the south, west and south, or west and north directions, respectively, and deletes if it corresponds to the 12 test templates.

If the weight is 3, the three adjacent black pixels are northwest and north and northeast, northwest and west and southwest, southwest and south and southeast, southeast and east and northeast, southwest and south and east, northwest and west and south, northeast and east. If applicable to nine test templates located in and south, west and south and southeast, or west and southwest and south, respectively, delete them, and delete them in north and northeast and east, south and southeast and east, or north and northwest and west, respectively. If three test templates are included, the deletion condition is deleted when there are one or more adjacent black pixels having a weight of three adjacent black pixels greater than seven.

If the weight is 4, four adjacent black pixels are northwest, north, northeast, east, southwest, west, northwest, north, west, northwest, north, northeast, south, southwest, west, northwest, southeast, south, southwest, and west. 10 test plates corresponding to east and south east and south and south west, north east and east and south east and south, north and north east and east and south east, north east and east and south and south west, or south east and south and west and north west respectively. Delete the case.

If the weight is 5, five adjacent black pixels are located in the east, southeast, south, north, northeast, east, west, northwest, north, south, southwest, west, northwest, north, northeast, or northeast, east, southeast. If it corresponds to six test templates each located in the excluded direction, delete it, and if it corresponds to two test templates located in the southwest and south and southeast, or in the direction except northwest and west and southwest, each of the adjacent black pixels with a weight of 8 Delete only if 1 or more.

If the weight is 6, six adjacent black pixels are deleted if they correspond to four test templates, each located in a direction other than northwest and north, east and southeast, northeast and east, or north and northeast. For each of the four test templates located in the west, southwest, west and northwest, or south and southeast directions, only the adjacent black pixels with a weight of 8 or more are deleted.

If the weight is 7, the seven adjacent black pixels correspond to the two test templates located in the direction except east or north, and the two test templates located in the direction except west or south are deleted. If applicable, it is deleted only when the adjacent black pixel having a weight of 8 is 2 or more.

If the weight is 8, it is deleted unconditionally.

Finally, if the weight is 9, it is deleted only in the first thinning step.

On the other hand, the present invention can be thinned without any problem in processing a two-line pixel image. 3A and 3B show a vertical pattern of two pixels and three pixels wide.

In the figures, each pixel (2,2) is an outer pixel, but it is not possible to know whether the vertical line pattern is three lines wide or two lines wide only by neighboring pixels of the 3x3 block. 6,3 and the pixels 6 and 4 of FIG. 3B likewise have no way of distinguishing the width of the vertical line pattern by only looking at 3x3 neighboring pixels.

However, when weighting is used, this difficulty can be easily solved by checking only neighboring pixels in a 3x3 block.

3C and 3D show weights of the input images of FIGS. 3A and 3B, respectively, to distinguish the two images.

In the figure, each pixel (2, 2) both has 3 as the weight. However, it can be seen that the weights of the pixels around the pixels are different from each other, and the weights of the pixels around 6 and 3 of FIG. 3C and the pixels 6 and 4 of FIG. 3D are different from each other. It can be seen.

Just because they have 8-neighbor pixels with different weights does not mean that all of the weights of these 8-neighbor pixels must be compared. However, if a black pixel has a neighboring pixel having a weight of 8 among the 8-neighbor pixels, it means that the black pixel is a neighboring pixel that can be deleted.

Meanwhile, in the process of repeatedly removing the outer pixel, an input image remains one pixel width or two lines thick. In the case of general image data, only even-width images exist. In this case, the test template and the deletion condition are formed according to the principle of deleting the right part in the vertical line and the top part in the horizontal line.

Those skilled in the art will appreciate that various changes and modifications can be made without departing from the spirit of the present invention.

As described above, according to the present invention, a method of thinning an outline of an input image, which can be widely used in various fields such as character recognition, defect inspection of a circuit board, fingerprint recognition, etc., simultaneously removing pixels satisfying a specific condition from an outline. By adopting the parallel thinning method, the weighting concept is introduced to achieve the improvement of efficiency, accuracy, and execution speed, while reducing the amount of information required for automatic recognition of image patterns to a minimum amount.

Claims (3)

A method of thinning outlines in binary image patterns, A first step of assigning a weight indicating a number of black pixels adjacent to each black pixel to a binary image pattern input to the image processing apparatus through an image input device; A second step corresponding to a test template of 3 × 3 blocks previously stored corresponding to a weight, and at the same time deleting the corresponding black pixel if the previously deleted condition is satisfied; Weight again until there is no black pixel that satisfies the deleting condition, and the black pixel deleting step is repeatedly performed to terminate the thinning process if there is no black pixel that satisfies the deleting condition. And a third step of outputting an image pattern. 2. The parallel thinning method according to claim 1, wherein in the third step, black pixels having weights indicating independent pixels without any black pixels adjacent to the black pixels are not deleted. The method according to claim 1 or 2, wherein the weight is assigned to 1 to 8 corresponding to the number of adjacent black pixels, and 9 if no adjacent black pixels are provided. If the weight is 1, there are eight test templates located in one adjacent black pixel east, west, south, north, northeast, northwest, southeast, or southwest direction, and the delete condition is that the weight of the adjacent black pixel is greater than three. Delete on, If the weight is 2, two adjacent black pixels are northeast and east, north and northeast, northwest and north, northwest and west, west and southwest, southwest and south, south and southeast, southeast and east, east and north, east and east It has 12 test templates located in the south, west and south, or west and north directions, and deletes if it corresponds to the 12 test templates. If the weight is 3, the three adjacent black pixels are northwest and north and northeast, northwest and west and southwest, southwest and south and southeast, southeast and east and northeast, southwest and south and east, northwest and west and south, northeast and east. If applicable to nine test templates located in and south, west and south and southeast, or west and southwest and south, respectively, delete them, and delete them in north and northeast and east, south and southeast and east, or north and northwest and west, respectively. If three test templates are included, the delete condition is deleted when there are one or more adjacent black pixels with weights of three adjacent black pixels greater than seven. If the weight is 4, four adjacent black pixels are northwest, north, northeast, east, southwest, west, northwest, north, west, northwest, north, northeast, south, southwest, west, northwest, southeast, south, southwest, and west. 10 test plates corresponding to east and south east and south and south west, north east and east and south east and south, north and north east and east and south east, north east and east and south and south west, or south east and south and west and north west respectively. Delete the case, If the weight is 5, five adjacent black pixels are located in the east, southeast, south, north, northeast, east, west, northwest, north, south, southwest, west, northwest, north, northeast, or northeast, east, southeast. If it corresponds to six test templates each located in the excluded direction, delete it, and if it corresponds to two test templates located in the southwest and south and southeast, or in the direction except northwest and west and southwest, each of the adjacent black pixels with a weight of 8 Only delete if more than 1, If the weight is 6, six adjacent black pixels are deleted if they correspond to four test templates, each located in a direction other than northwest and north, east and southeast, northeast and east, or north and northeast. , If the four test templates are located in each of the west and southwest, west and northwest, or south and southeast directions, only the adjacent black pixels with a weight of 8 or more are deleted. If the weight is 7, the seven adjacent black pixels correspond to the two test templates located in the direction except east or north, and the two test templates located in the direction except west or south are deleted. If applicable, delete it only if adjacent black pixels with a weight of 8 are 2 or more, If the weight is 8, it is deleted unconditionally. If the weight is 9, the parallel thinning method using weights, which is deleted only in the first thinning step.