JP2006244083A - Image processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus that produces a clear image by erasing only hole portions on an image acquired by scanning document with binder holes. <P>SOLUTION: The image processing apparatus comprises an image reading means 7 for reading image information on a document, an image storage part 50, a low pass filter means applied to the image read by the image reading means 7, a feature detection means for detecting features of blocks of continuous image areas in the image read by the image reading means 7, a document hole detection means for detecting holes in the document according to the features detected by the feature detection means, a hole image creation means for leaving only the holes detected by the document hole detection means on the image read by the image reading means 7, and a hole erasing means for erasing the hole portions on the image read by the image reading means 7 according to the image created by the hole image creation means. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus that processes image information of a read original.

  In general, an image reading device is a device that reads a document such as a report, a minutes, a design drawing, and a confidential document, and stores it as image data. In addition to the image reading function, image processing functions such as blank page skip and OCR (optical character reader) have been devised. On the other hand, when a binder hole is opened in the document for filing to the binder, the binder hole portion of the read image becomes a black dot. This is because the opposing surface of the image reading portion of the image reading apparatus is black. The binder hole portion in this image causes blank page skip and OCR to malfunction.

  B) Here, the influence of a document having a binder hole on the OCR function will be described.

  FIG. 2 shows an example in which the scanned image (a) obtained by reading the dotted line area of the document (b) with the image reading apparatus is subjected to OCR. Here, if the document has a binder hole, an image of the hole portion is formed in the scanned image (a). (Hereinafter, the image of the binder hole portion is referred to as a binder hole image.) When this scanned image (a) is subjected to OCR, the binder hole portion is recognized as “•”, and as a whole, as shown in FIG. " This is a misrecognition.

  B) Next, the effect on blank page skip will be described.

  As a known technique for skipping blank pages, there is a technique of “detecting blank pages based on the ratio of black pixels occupying the image area and discarding the detected read images of blank pages”. A specific example is shown in FIG. Here, the ratio of black pixels is set to 2%. During the continuous scanning of the first, second, and third sheets, if there is a blank document on the second sheet, the percentage of black pixels in the scanned image is 0%. Discarded by blank page skip processing. However, as shown in FIG. 3B, when the document has a binder hole (here, the proportion of black pixels is 5%), the second image has a binder hole image. Since this is possible, this image is not judged to be blank and is not discarded. When the proportion of black pixels is increased to the proportion of the image of the binder hole, there is a possibility of discarding the image that is desired to be left, so that there is a problem that the reliability of blank page skip is lowered.

  In order to avoid the above-described problems of OCR and blank page skip, a technique for deleting the binder hole image has been considered.

  As a technique for erasing the binder hole image, there is a method of “erasing the image area specified by the user” as shown in FIG. In FIG. 8A and FIG. 8B, the area from the document edge to the dotted line is painted white. The user can specify the width from the document edge to the dotted line.

  Further, there is a method of changing the facing surface of the image reading portion of the image reading apparatus from black to white. This prevents binder holes from appearing in the image.

In addition, there is a method in which black pixels are counted in a set partial area, and a binder hole is used as noise based on the count value (see, for example, Patent Document 1).
Japanese Patent Laid-Open No. 2-68676

  However, in the case of “deleting the image area specified by the user”, there is a problem that, as shown in FIG. Specifically, FIG. 8 (a) - for In the case of (c), the erase region indicated by the dotted line, the "Hello", "this" in the sentence have overlapped bore portion and a vertical line, the hole When erasing part of the image, it disappears together. Similarly, when the area indicated by the dotted line is erased, the upper part of “<1>” disappears in FIGS. 8B to 8D for the same reason.

  In addition, in the “method of changing the facing surface of the image reading portion of the image reading apparatus from black to white”, the size of the original in the image is also whitened, so that the size detection process cannot be performed. Here, the size detection process will be described with reference to FIG. FIG. 6A shows an original, and FIG. 6B shows an image obtained by reading the original shown in FIG. In this image reading apparatus, since the facing surface of the image reading portion is black, the area around the original in the read image becomes black as shown by the oblique lines. The image (b) is examined from around and the position of the original pixel is detected by detecting the position where the black pixel is changed to the white pixel. When performing the size detection process by such a method, there is a problem that the size detection process cannot be used if the above-described “method of changing the facing surface of the image reading portion of the image reading apparatus from black to white” is used.

  Moreover, in patent document 1, when the character 605 of FIG. 4 is covering the binder hole 604, a binder hole cannot be erased.

  Therefore, the present invention has been made to solve the above-mentioned problem, and the first object is to form a binder hole even when the characters 605 and the table 602 are covered with the binder holes 601 and 604 as shown in FIG. An image processing apparatus capable of erasing the image is provided.

  In order to solve the above object, the present invention comprises the following arrangement.

  Image reading means for reading image information of a document, an image storage unit, low-pass filter means for processing a working image that is a copy of the read image read by the image reading means, and the work processed by the low-pass filter means A feature amount detecting means for detecting a feature amount of a continuous block of image regions in the image for use, a document hole detecting means for detecting an image of a hole in the document based on the feature amount detected by the feature amount detecting means, Based on the image created by the hole image creating means, the hole image creating means for erasing other than the hole image detected by the original hole detecting means from the working image processed by the low pass filter means, An image processing apparatus comprising hole erasing means for erasing a hole portion of an image read by an image reading means.

  According to the present invention, only the binder hole image can be erased regardless of the position and shape. Also, unlike the “method of erasing the image area specified by the user”, the area other than the binder hole is not erased.

  Further, the appearance of the read image is improved by the disappearance of the binder hole. In addition, the recognition rate of OCR is improved.

  In addition, it is possible to contribute to prevention of malfunctions in the blank page skip processing, and it is possible to erase unnecessary information by erasing the binder hole image, thereby contributing to increasing the compression rate of image compression. Furthermore, even when the binder hole image is covered with a front line or when the binder hole image is covered with another image, only the binder hole can be erased. When a scanner having a black sensor facing surface is used, the size detection process and the binder hole erasing process can be used together.

  The best mode for carrying out the present invention will be described below based on examples.

  Examples of the present invention will be described below. FIG. 1 is an external view of an image reading apparatus (hereinafter referred to as a scanner) that reads an image. The read original is fed from the paper feed unit 1 and discharged from the paper discharge unit 2. FIG. 5 is an explanatory diagram of the flow of reading the surface of the document. The document 3 is conveyed along the conveyance direction 6, and the surface image of the front surface 4 is read by the reading unit 7 that is the image reading unit of claim 1, and the image storage unit that is the image storage unit of claim 1. 50. The image storage unit is a memory in the scanner.

  Next, the feature amount detection means of claim 1 will be described. Assume that the image is binarized. The types of feature quantities detected by the feature quantity detection means include the perimeter, the position of the center of gravity, the area, and the circularity. The method of obtaining the perimeter of the in-region image and the center of gravity uses contour line tracking processing. FIG. 12 is an explanatory diagram of the contour line tracking process. First, the image is scanned from the upper left corner toward the lower right corner to determine a black pixel that is a starting point for contour tracking (FIG. 12-1). When the starting point is determined, the vicinity of the point is examined counterclockwise (FIG. 12-2). Here, when the neighborhood of 8 is examined, if the next black pixel does not exist, the start point is an isolated point, the process is stopped, and scanning is performed up to the start point of the next in-region image (FIG. 12-3). If it is not an isolated point, the black pixel that appears first is set as the next new tracking point, and the vicinity of the tracking point is examined in the counterclockwise direction (FIG. 12-4). When the scanning in FIG. 12-4 is repeated, one contour line can be obtained. At this time, the number of pixels of the contour line is set as the peripheral length (* corrected by multiplying by √2 with respect to the length in the oblique direction. See FIG. 14), and the total value of the x coordinate and y coordinate of the contour line is The center of gravity is obtained by dividing the number of pixels of the contour line.

The method for obtaining the area is obtained by using the following filling process. First, the painting process will be described with reference to FIG. First, assume that the area is filled as shown in FIG. First, a starting point ((X, Y)) is determined, and the filling process is started from that point. Scan from the start point (X, Y) in the left-right direction to find the boundary of filling. The boundary condition is (1) the edge of the document. (2) When the color changes. There are two. (The X coordinates of the obtained boundary are assumed to be X _R and X _L , respectively) (Process 1). Next, the line segment (X _R , Y)-(X _L , Y) is painted (process 2). Next, (X _R , Y + 1) − (X _L , Y + 1) is scanned from the left to find all the coordinates of both ends of the line segment to be filled and stored in the buffer (processing 3) (specifically, The line segments (X _R , Y + 1)-(X _L , Y + 1) are scanned to search for line segments to be filled in. Here, line segments ab and cc are found (where c− (c is only one pixel.) The coordinates of the end points of this line segment are stored in the buffer.) Next, the same processing is performed for (X _R , Y-1)-(X _L , Y-1) (processing 4). Thereafter, the coordinates stored in the buffer are taken out and processing 1 to processing 4 are repeated. The fill process is complete when the buffer is empty. For the area, the number of filled pixels is counted, and the number of pixels is defined as the area.

  Next, how to determine the circularity will be described. Circularity can be obtained by applying the perimeter and area to the following formula.

D = (4π × Area) / (Length × Length)
Here, D = circularity. Area = Area. Length = perimeter length.

  When this D becomes 1, the image becomes a circle. In Example 1, the feature amount of the binder hole is 0.8 or more.

  Next, the entire flow of the binder hole image erasing process will be described with reference to the flowchart of FIG. First, image information is read by the scanner 100 for a document having a binder hole (S901). The read surface image is stored in the image storage unit 50 (S902) and converted into a binary image (S903). Next, a copy of the binary surface image is created and stored in the image storage unit (S904). Hereinafter, this image is referred to as a work image. Next, a low-pass filter process described later is performed on the work image (S905). Next, the feature amount of a continuous image area in the image is checked with respect to the work image, and the binder hole portion is detected (S906). Next, an image of only the detected binder hole portion is created (S907). Hereinafter, the image of only the binder hole is referred to as a hole image. Next, the binder hole of the original image is deleted based on the hole image (S908). Details of the deletion will be described later.

  Next, a low pass filter according to the low pass filter means of claim 1 will be described. FIG. 7 is a diagram for explaining the operation of the low-pass filter means. First, the low pass filter means scans in the main scanning direction from the upper left end of the image toward the lower right end. At that time, the number of continuous black pixels is counted, and the threshold value T is compared with the number of black pixels n counted when the color of the target pixel becomes white. In the case of n <T, the pixels counted up to now are painted white. Then, n is returned to zero. If n ≧ T, return n to zero. The above scanning is performed to the lower right end. Further, the same processing is performed in the sub-scanning direction. With the above processing, an image whose width and height are smaller than T in the original image is deleted. By this low-pass filter processing, it is possible to erase the table and characters that are covered with the binder hole image as shown in FIG.

  Next, a hole image creating method according to the hole image creating means of claim 1 will be described with reference to FIG. FIG. 10 is a flowchart for creating an image of only the binder hole from the working image, and is a detailed explanatory diagram of steps S906 and S907 in FIG. First, the work image is scanned from the upper left end to the lower right end (S1001). In S1002, it is determined whether or not the pixel of interest is at the lower right end. If it is at the lower right end, the process proceeds to S1007, and the process ends. Otherwise, the process proceeds to S1003, and it is determined whether or not the target pixel is a black pixel. If the target pixel is not a black pixel, the process returns to S1001. If the target pixel is a black pixel, a feature amount of a continuous image area including the target pixel is extracted (S1004). Next, in S1005, when the extracted feature amount satisfies the condition value for determining that it is a binder hole, the process returns to S1001 and scanning is continued. If not satisfied, a group of image areas including the target pixel is not a binder hole, and the image area is deleted by white filling processing (S1006). Thereafter, the above-described processing is continued until the target pixel becomes the lower right corner (Yes in S1002). When the processing is completed (S1007), an image (hole image) in which only the binder hole remains is created.

  Next, an example of a condition value for determining that the extracted feature amount is that of the binder hole is shown in the table.

  When this table is applied, it is considered as a binder hole. The processing for determining the binder hole based on the above table is related to the document hole detecting means of claim 1.

  Next, the processing for erasing the binder holes in the original image according to the hole erasing means of claim 1 will be described with reference to FIG. The image in FIG. 11A is an original image. The image in FIG. 11B is a hole image. The two images are overlapped, and the portion where the black pixels overlap is whitened. If the original image is not binarized, the portion of the original image corresponding to the hole image is whitened or replaced with a surrounding approximate color.

  The binder hole eliminating process can be realized by the above process.

  In consideration of the possibility that the size of the hole image changes due to the low-pass filter processing, the region to be subjected to the hole erasing processing may be set to be slightly wider or narrower than the hole image.

  As the low-pass filter means, in addition to those performing the above operation, various means having a function capable of removing fine lines such as a weighted average of surrounding pixels can be used.

  If only a binarized image is required as an output image, the binarizing means binarizes the original image, and then a copy thereof may be used as a work image. Further, the binarization threshold value of the original image and the work image may be changed.

  When the low-pass filter means is applicable to a multi-valued image, the work image that has undergone the low-pass filter processing may be binarized.

  Also, it is within the scope of the present invention to consider a scanner and an information processing device such as a PC connected thereto as an image processing apparatus. In this case, the image storage unit may be provided only on the information processing device side, and the binarizing unit, the low-pass filter unit, the feature amount detecting unit, the document hole detecting unit, the hole image creating unit, and the hole erasing unit It may be provided on the side. Further, the binarizing means, the low-pass filter means, the feature amount detecting means, the document hole detecting means, the hole image creating means, and the hole erasing means may be realized as functions of a computer program that operates on the information processing device.

  Further, the image reading apparatus is also provided with a function equivalent to that of the information processing device, and the binarizing means, the low-pass filter means, the feature amount detecting means, the document hole detecting means, the hole image creating means, and the hole erasing means are included in the image reading apparatus. The control circuit may be implemented as a computer program function that operates in the image reading apparatus.

External perspective view of the scanner Illustration of the impact of OCR recognition Illustration of the effect of skipping blank pages Illustration when the front and letters are covered in the binder hole Outline diagram about transportation Size detection explanation diagram Low pass filter illustration Explanatory drawing of an example to erase the binder hole by specifying the area Flowchart of explanation 1 of binder hole erasing process Flowchart of explanation 2 of binder hole erasing process Illustration of how to use hole images Illustration of contour tracking Explanatory drawing about the filling process Illustration of pixel and perimeter

Explanation of symbols

7 Image reading unit (corresponding to image reading means)
50 Image Storage Unit 100 Scanner

Claims (4)

  Image reading means for reading image information of a document, an image storage unit, low-pass filter means for processing a working image that is a copy of the read image read by the image reading means, and the work processed by the low-pass filter means A feature amount detecting means for detecting a feature amount of a continuous block of image regions in the image for use, a document hole detecting means for detecting an image of a hole in the document based on the feature amount detected by the feature amount detecting means, Based on the image created by the hole image creating means, the hole image creating means for erasing other than the hole image detected by the original hole detecting means from the working image processed by the low pass filter means, An image processing apparatus comprising hole erasing means for erasing a hole portion of an image read by an image reading means.   2. The image processing apparatus according to claim 1, further comprising: binarizing means for binarizing the multi-value data, wherein the read image and the work image are binarized images.   The binarizing means for binarizing the multi-value data is added, and the work image is binarized by the binarizing means and then processed by the low-pass filter means. Image processing apparatus.   2. The binarization means for binarizing multi-value data is added, and after the working image is processed by the low-pass filter means, the binarization means binarizes. Image processing apparatus.

Images