KR100205599B1 - How to Binarize Text Documents - Google Patents

Abstract

end. The technical field to which the invention described in the claims belongs

A method of binarizing an image obtained by scanning a document in an image input apparatus.

I. The technical problem to be solved by the invention

Provided are a binarization method of a character document image that can improve image quality and compression ratio even for an image having a noise component.

All. Summary of Solution of the Invention

If the pixel of interest in the scanned image is the boundary of the character, it is examined for continuity according to the directionality. If there is continuity, it is binarized to the same value as the binarization value for the previous pixel in the same direction. Apply and binarize.

la. Important uses of the invention

Used to binarize text document images.

Description

How to Binarize Text Documents

The present invention relates to an image input device, and more particularly, to a method for binarizing an image obtained by scanning a document.

BACKGROUND ART Image input devices employed in facsimile, scanner, digital copier, and the like are typically processed after converting an analog image signal obtained by scanning a document into digital image data. At this time, the analog image signal is converted to digital image data having a gradation corresponding to the original document for each pixel. Since the gradation has a gradation, the amount of data becomes very large. For example, in the case of 256 gray levels, each pixel is represented by 8 bits. Accordingly, by binarizing the image data according to the scan, the amount of data is drastically reduced and then encoded and transmitted or a copy document is output. The binarization determines each pixel having a gradation as a "white" pixel or a "black" pixel, and the "white" pixel and the "black" pixel are relatively represented by 1-bit data of "1" or "0".

This binarization adopts a different method depending on the type of document image. The document image is classified into a text document consisting of text, a picture document consisting of a picture (or a picture), and a document in which text and pictures are mixed. Among these, the method of binarizing characters is a method of using a global threshold and a method of using a local threshold. Among them, the method using the local threshold reflects the characteristics of the document better, so the result is better. When applying a local threshold, this may be done for the entire area of the image or only at the edge part.

The typical character binarization method that has been used before is the maximum and minimum values of the surrounding pixel values of the pixel of interest, which are the pixels to be binarized, and if the difference is greater than the preset threshold, the edge is recognized as a local threshold. If it is smaller than the threshold, it is recognized as a background and binarized as a global threshold. This method is commonly referred to as a minimum-maximum method. This method combines the advantages of using a global threshold and a local threshold, and shows good results in general documentation.

However, a problem in the above-described minimum-maximum method is that since edges are detected without considering the directionality and continuity of the edges, the noise components are extracted, not the edges of the actual character. As a result of this, a bad result is obtained for an image in which a noise component exists, and there is a disadvantage in that the extracted edge is also not straightened. As such, if the edge is not straightened, the compression rate is lowered during data compression coding such as run-length coding, so that the transmission efficiency is lowered and the image quality is degraded during data transmission.

As described above, the conventional method of binarizing a text document image detects and binarizes edges without considering the directionality and continuity of the edges. There was a downside to this falling.

Accordingly, an object of the present invention is to provide a binarization method of a character document image that can improve image quality and compression ratio even for an image having a noise component.

1 is a block diagram of a conventional scanner to which the present invention is applied;

2 is a process flow diagram in accordance with an embodiment of the present invention.

3 is an exemplary window for detecting the size and direction of a typical edge;

Figure 4 is an exemplary view showing a method for investigating continuity according to the direction in the present invention,

5 is an exemplary view showing a method for determining a local threshold in the present invention,

According to the present invention for achieving the above object, if the pixel of interest in the scanned image is the boundary portion of the character, the continuity according to the directionality is examined, and if there is continuity, it is binarized to the same value as the binarization value for the previous pixel in the same direction. And if there is no continuity, it is binarized by applying a local threshold.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and the annexed drawings, numerous specific details are set forth in order to provide a more thorough understanding of the present invention, such as specific processing flows or windows. It will be apparent to those skilled in the art that the present invention may be practiced without these specific details. And a detailed description of known functions and configurations that may unnecessarily obscure the subject matter of the present invention will be omitted.

Figure 1 shows a block diagram of a conventional scanner to which the present invention is applied. The central processing unit 100 detects the document mounting by the document detection sensor 112 and controls the image processor 104 to scan the document when a scan start command is input from the operation panel 110. The scanning unit 102 includes an optical system for scanning an original and scans the original to generate an analog image signal. The image processor 104 drives the scan unit 102 to convert the image signal output from the scan unit 102 into digital data, and to process and binarize the image. The ROM (Read Only Memory) 106 stores the operation program of the CPU 100 and various reference data in advance. The random access memory (RAM) 108 temporarily stores data according to an operation of the CPU 100. The operation panel 110 provides a key input by the user to the CPU 100. The document detection sensor 112 detects that the document is mounted and notifies the central processing unit 100.

The present invention binarizes the edge portion in consideration of the directionality and continuity of the edge in order to solve the above-mentioned disadvantages in the image input apparatus such as the scanner. In other words, if the continuity of the edge is present, the binarization result is straightened.

2 is a flowchart illustrating a process of binarizing one pixel for convenience, according to an exemplary embodiment of the present invention. The same applies to all pixels. The operation according to the flowchart of FIG. 2 is programmed to be performed by the image processor 104 of FIG.

As the document is detected by the document detection sensor 112 and the scan start command is input from the operation panel 110 in the above state, the image processor 104 is controlled by the central processing unit 100 under the control of the scanning unit ( When the original is scanned by 102, an image signal corresponding to the original scan is output from the scanning unit 102. Then, the image processor 104 converts the image signal output from the scan unit 102 into digital image data, stores the pixel value in the RAM 108 in units of lines scanned, and binarizes according to the flowchart of FIG. 2. do. In the following description, it is assumed that the number of lines for storing the scanned image is based on three lines as a default.

First, the image processor 104 designates the pixel of interest to be binarized in operation 200 of FIG. 2, and then detects the size and direction of the edge in operation 202. At this time, as shown in Figs. 3A to 3D, the size and the direction of the edge are detected using four windows that can determine the direction. FIG. 3 (a) shows a window in the direction of 0 °, FIG. 3 (b) shows a window in the direction of 45 °, and FIG. 3 (c) shows a window in the direction of 90 °, and FIG. 3 (d). ) Shows the window at 135˚. 3 (a) to 3 (d) show three lines, i.e., the pixel of column j of the i-th line as the pixel of interest, the i-1th line, the i-th line and the i + 1th line, respectively. By giving the corresponding weights to the pixels of the j-1 column, the j column, and the j + 1 column, the edge detection method is a known Kirsch method. Therefore, detailed description thereof will be omitted.

After detecting the size and orientation of the edge by the edge detection method using four windows as described above, the image processor 104 determines the size of the edge, that is, four size values obtained by using four windows in step 204. Value is compared with the preset threshold. If the size of the edge is larger than the threshold, the pixel of interest is the boundary of the character. If the edge is smaller than the threshold, the pixel of interest is the inside or background of the character, not the boundary.

At this time, if the size of the edge is not larger than the threshold value, the pixel of interest may be binarized in the same way as the conventional minimum-maximum method because the pixel of interest is not the boundary of the character but the inside or the background. Accordingly, the image processor 104 binarizes the global thresholds in steps 206 to 210 as in the min-max method. That is, the image processor 104 compares the pixel of interest value with the global threshold in step 206, binarizes the pixel of interest to “white” if it is greater than the global threshold in step 208 and 210 if it is smaller than the global threshold in step 208. The pixel of interest is binarized to "black." At this time, the global threshold value is set to, for example, half of the maximum gradation value of the input image.

On the other hand, if the size of the edge is larger than the threshold in step 204, the pixel of interest may be viewed as the boundary of the character. However, in order to prevent the deterioration of the image quality or the straightening of the edge due to noise components (212). In step 222, the edge portion is binarized in consideration of the directionality and continuity of the edge.

In step 212, the image processor 104 determines whether the image processor 104 is continuity with a previously detected edge according to the direction of the edge detected in step 202. At this time, according to the directionality of the edge determined in one direction among 0 °, 45 °, 90 °, and 135 ° by the windows of FIGS. 3A to 3D as described above, FIGS. 4A to 4D. Examine the continuity by applying one of them. 4 (a) to 4 (d) show that the pixel of interest detected as an edge is examined for continuity with the shaded pixel. That is, when it is detected that the pixel of interest is an edge having a directionality of 0 °, it is checked whether there is an edge of 0 ° like the pixel of interest in the hatched portion as shown in Fig. 4A. Similarly, when the pixel of interest is detected to be an edge having a directionality of 45 °, it is checked whether the pixel of interest has an edge of 45 ° like the pixel of interest as shown in FIG. 4 (b). If it is detected that the branch is an edge, as shown in Fig. 4 (c), it is examined whether there is an edge of 90 degrees, such as the pixel of interest, and if it is detected that the pixel of interest is an edge having directionality of 135 degrees, Fig. 4 Investigate whether there are 135 ° edges, such as the pixel of interest, in the hatched areas as shown in (d). By examining the continuity in this way, the edge can be straightened.

After checking the continuity in step 212, the image processor 104 checks whether the irradiation result is continuity in step 214. If there is continuity, the edge will be straightened if the pixel of interest is binarized with the previous pixel in the same direction. In order to straighten the edges, the image processor 104 binarizes the pixel of interest to the same value as the previous pixel in the same direction in step 216. At this time, the previous pixel becomes the hatched pixel in one of Figs. 4 (a) to 4 (d) according to the direction of the pixel of interest.

On the other hand, if there is no continuity in step 214, it is not necessary to consider the previous pixel with respect to the pixel of interest, so that it is binarized by applying a local threshold as in the conventional minimum-maximum method. Accordingly, the image processor 104 applies and localizes local thresholds as in the minimum-maximum method in steps 218 to 222. That is, the image processor 104 compares the pixel value of interest with the local threshold in step 218, binarizes the pixel of interest to “white” in step 220 if it is greater than the local threshold, and in step 222 if it is smaller than the local threshold. The pixel of interest is binarized to "black."

Herein, a method of determining the local threshold is shown in Figs. 5 (a) to 5 (d). FIG. 5A shows that when the edge of the pixel of interest has a direction of 0 °, the average value of the hatched portion is set as the local threshold. Similarly, in FIGS. 5B, 5C, and 5D, when the edges of the pixel of interest have directivity of 45 °, 90 °, and 135 °, the average value of the hatched portion is set to the local threshold. see. At this time, the portion marked with "▩" may be given a weight.

When the above-mentioned binarization process for one pixel is finished, the same operation is repeated until the next pixel is designated and binarized again until the binarization of the entire image is completed.

Therefore, since the binarization is performed in consideration of the directionality and continuity of the edge, straightening of the edge can be achieved and generation of an isolated point due to noise can be prevented. Therefore, only the edges of the actual characters can be excellently extracted even for an image having a noise component.

Meanwhile, in the above description of the present invention, specific embodiments have been described, but various modifications can be made without departing from the scope of the present invention. In particular, the method of determining the global threshold or the local threshold shown in the embodiment of the present invention may vary as needed. Therefore, the scope of the invention should not be defined by the described embodiments, but should be defined by the equivalents of the claims and the claims.

As described above, in the binarization method of a character document image, the edge continuity is linearized when the continuity of edges is present. In addition, the image quality can be improved by extracting only the edges of the actual characters with respect to an image having a noise component.

Claims (8)

In the method for binarizing an image obtained by scanning a text document, If the pixel of interest of the scanned image is the boundary of the character, the continuity according to the direction is examined. If there is continuity, the pixel is binarized to the same value as the binarization value for the previous pixel in the same direction. Binarization method characterized in that the binarization by applying. In the method for binarizing an image obtained by scanning a text document, A size and directionality detection process of detecting the size and directionality of an edge of the pixel of interest of the scanned image; Comparing the size of the edge with a preset threshold to determine whether the pixel of interest is a boundary of a character; A first binarization process of binarizing the pixel of interest by applying a global threshold when the pixel of interest is detected to be not the boundary; A continuity investigation process for investigating whether the pixel of interest is the boundary portion and whether the edge according to the orientation is continuity; A second binarization process of binarizing the pixel of interest to a value equal to a binarization value for the previous pixel in the same direction when the edge is continuous; And a third binarizing process for binarizing the pixel of interest by applying a local threshold when there is no continuity of the edge. The binarization method according to claim 2, wherein the detecting of the size and the directionality detects the size and the directionality in four directions of 0 °, 45 °, 90 °, and 135 ° with respect to the pixel of interest. 3. The method of claim 2, wherein the continuity checking process checks the previous pixel in the same direction as the directionality of the pixel of interest and detects that there is continuity if the previous pixel is the same directional edge, otherwise there is no continuity. Binarization method characterized by the above-mentioned. 5. The method of claim 4, wherein the first binarization process compares the pixel of interest value with the global threshold, and if the pixel of interest value is greater than the global threshold, binarizes the pixel of interest to a “white” pixel and is greater than the global threshold. If smaller, the pixel of interest is binarized to "black" pixels. 6. The method of claim 5, wherein the third binarization step compares the pixel value of interest with the local threshold, and if the pixel value of interest is greater than the local threshold, binarizes the pixel of interest to a " white " If smaller, the pixel of interest is binarized to "black" pixels. 7. The binarization method according to claim 6, wherein the global threshold is set to half of the maximum gradation value of the input image. 8. The binarization method according to claim 7, wherein the local threshold is set by selectively taking an average value with respect to adjacent pixels according to the directionality of the pixel of interest.

Images