JP2006268371A - Image processor and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To assume that difference of states of respective partial images to an original image exists, and to speed up processing for retrieving a plurality of common characteristic points, in an image processor and an image processing method joining the plurality of partial images obtained by dividing the original image into a plurality of pieces and reading them. <P>SOLUTION: The original image is divided into the plurality of pieces and is read such that one portion is overlapped, the plurality of pieces of partial read image data expressing the plurality of partial images are generated (100-105), a slant angle and an erect direction of each the partial image to the original image are detected (110, 112), the plurality of characteristic points in each the partial image are extracted by use of a filter corresponding to the detected slant angle (116), a set of the partial images having the plurality of common characteristic points is retrieved from the plurality of partial images on the basis of the extracted characteristic points (117-119), the retrieved set of the partial images is joined such that the plurality of characteristic points are overlapped, and original image data expressing the original image are generated (120). <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image processing apparatus and an image processing method, and more particularly to an image processing apparatus and an image processing method for joining a plurality of partial images obtained by dividing an original image into a plurality of parts.

  Conventionally, when joining a plurality of partial images obtained by changing the reading position so that a part of the original image overlaps a plurality of times, the feature points are extracted from the partial images and from the adjacent partial images. Two common feature points are searched, and a plurality of partial images are joined so that the two common feature points overlap.

  In addition, since each of the partial images may be inclined with respect to the original image or may have different erecting directions, it is assumed that there are differences in the state of the partial image with respect to the original image, such as inclination and direction. Then, the number of types of feature points extracted from the partial image is increased, or two common feature points are searched by expanding the search range while rotating the vector direction based on the positions of the two feature points.

  In an apparatus for joining images, a point on the edge of one image is extracted as a feature point, and a correlation between each feature point and a point on the edge in the other image is obtained. An image synthesizing apparatus that synthesizes image data while determining a corresponding area of an image and deforming so as to cancel the shift is known (Patent Document 1).

Also, by calculating the angle of rotation of one partial image with respect to the other partial image based on the two feature points captured in the image obtained by applying the low pass filter to the partial image, An image synthesizing method for synthesizing while rotating is known (Patent Document 2).
JP-A-11-015951 JP 2000-078467

  However, in the image synthesizing apparatus described in Patent Document 1, assuming that there is a difference in state such as inclination and direction of the image with respect to the original image, the correlation between the feature point of one image and the other image is obtained. Since the number of points increases, there is a problem that it takes time for the correlation calculation processing for each feature point to determine the corresponding region. Further, in the image composition method described in Patent Document 2, assuming that there is a difference in state such as inclination and direction of the partial image with respect to the original image, it takes time to search for two captured feature points. There is a problem of this.

  The present invention has been made to solve the above-described problems, and assumes that there is a difference in the state of each partial image with respect to the original image, and that high-speed processing for searching a plurality of common feature points is performed. It is an object of the present invention to provide an image processing apparatus and an image processing method that can be realized.

  In order to achieve the above object, an image processing apparatus according to the present invention is represented by each of a plurality of partial read image data obtained by reading an original image in a plurality of times so that a part thereof is overlapped. Detection means for detecting information indicating a state of each partial image with respect to the original image; and a plurality of feature points common to the plurality of partial images based on the detected information and a plurality of feature points in each of the partial images. A search means for searching for a set of partial images having a combination of the searched partial images so that the plurality of common feature points overlap, and generating original image data representing the original image Generating means.

  Further, the image processing method according to the present invention provides the original image of each of the partial images represented by each of the plurality of partial read image data obtained by dividing the original image into a plurality of times and reading so that a part thereof overlaps. Detecting information indicating a state with respect to, and searching a set of partial images having a plurality of common feature points from the plurality of partial images based on the detected information and a plurality of feature points in each of the partial images, The set of the searched partial images is joined so that the plurality of common feature points overlap, and original image data representing the original image is generated.

  According to the present invention, the information indicating the state of each of the partial images represented by each of the plurality of partial read image data obtained by dividing the original image into a plurality of times so as to partially overlap the original image is obtained. And detecting a set of partial images having a plurality of common feature points from the plurality of partial images based on the detected information and a plurality of feature points in each of the partial images. A plurality of common feature points are joined so as to overlap, and original image data representing the original image is generated.

  Accordingly, by detecting information indicating the state of each partial image with respect to the original image, it is assumed that there is a difference in the state of the partial image in the range in which a plurality of feature points common to the set of partial images to be joined are searched. Therefore, since it is not necessary to enlarge, it is possible to speed up the process of searching for a plurality of common feature points.

  Further, the detection means uses the inclination angle of each partial image with reference to a predetermined direction of the original image, the erect direction of each partial image with respect to the original image, and each partial image with respect to the original image as information indicating the state with respect to the original image. At least one of the positions can be detected. As a result, even if the inclination angle and the erecting direction of each partial image are different, there is no need to expand the range for searching for a plurality of feature points common to the set of partial images to be joined, and the partial image relative to the original image By detecting each position, it is possible to specify a range for searching for a plurality of common feature points, and to speed up the process of searching for a plurality of common feature points.

  Further, the search means extracts a plurality of feature points in each partial image by processing a plurality of partial read image data using a filter corresponding to the detected information, and detects the detected information and the extracted portion. Based on a plurality of feature points in each image, a set of partial images having a plurality of common feature points can be searched from a plurality of partial images. Accordingly, it is possible to prevent an erroneous feature point from being searched for as a plurality of common feature points due to a difference in state of the partial image with respect to the original image, so that the accuracy of the joining process can be improved.

  In addition, the search unit corrects the plurality of partial read image data based on the detected information so that the states of the plurality of partial images with respect to the original image match, and then based on the corrected plurality of partial read image data. Then, a plurality of feature points in each partial image is extracted, and a set of partial images having a plurality of common feature points is searched from the plurality of partial images based on the plurality of feature points in each of the extracted partial images. be able to. By correcting a plurality of partial read image data so that the states of the plurality of partial images with respect to the original image coincide with each other, a range for searching for a plurality of feature points common to a set of partial images to be joined is set to a state of the partial image. Therefore, it is not necessary to expand assuming that there is a difference between them, and therefore, it is possible to speed up the process of searching for a plurality of common feature points.

  Further, the detection means detects at least one of an inclination angle of each partial image with respect to a predetermined direction of the original image as information indicating a state with respect to the original image, and an erecting direction of each partial image with respect to the original image, The search means rotates a plurality of partial read image data so that at least one of an inclination angle of each partial image with reference to a predetermined direction of the detected original image and an erect direction of each partial image with respect to the original image coincide with each other. After the correction, a plurality of feature points in each of the partial images are extracted based on the plurality of partial read image data subjected to the rotation correction, and from the plurality of partial images based on the plurality of feature points in each of the extracted partial images. A set of partial images having a plurality of common feature points can be searched. As a result, even if the inclination angles and erecting directions of the partial images are different, there is no need to expand the range for searching for a plurality of feature points common to the set of partial images to be joined. Can be speeded up.

  As described above, according to the image processing apparatus and the image processing method of the present invention, there is a difference in the state of each partial image with respect to the original image by detecting information indicating the state of each partial image with respect to the original image. As a result, it is possible to speed up the process of searching for a plurality of common feature points.

  Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. In the first embodiment of the present invention, the present invention is applied to a scanner device.

  As shown in FIG. 1, an image processing system 10 reads an image formed on a document and performs various image processing, and a scanner device 12 according to the first embodiment of the present invention, and image creation, processing, editing, and the like. The scanner apparatus 12 and the client PC 14 are connected by a network 16 such as a LAN.

  The scanner device 12 includes a ROM 20 that stores various programs and parameters including a read image joining processing routine program, which will be described later, a CPU 22 that executes various programs, and a RAM 24 that is used as a work area when the CPU 22 executes various programs. An image reading unit 26 that reads an original image formed on a document by scanning an provided image sensor (not shown) and generates partial read image data; an HDD 28 for storing image data; a network interface 30; And a bus 32 for connecting them to each other. In addition, the scanner device 12 has a conventionally known general function, and a description regarding the reading function is omitted.

  The client PC 14 is provided with a driver (not shown) for setting instructions (hereinafter referred to as “reading settings”) for various processing functions provided in the scanner device 12. The client PC 14 only needs to have a general configuration of a conventionally known personal computer, and the description of the general processing of the client PC 14 is omitted in the present embodiment.

  Next, the read image joining processing routine of the first embodiment will be described with reference to FIG. In the first embodiment, a case where an original image is scanned in three steps will be described as an example.

The scanner device 12 first determines in step 100 whether or not a scan instruction has been issued from the client PC 14. A part of the document representing the original image is set in the image reading unit 26 of the scanner device 12 by the user so that either the height direction or the width direction of the original image represented by the document is equal to the sub-scanning direction of the scanner device. is, the scan instruction from the client PC14 is input, the process proceeds from step 100 to step 101, in step 101, the scanning process is performed, the image data a ₁ reading portion is color image data is generated. In step 102, it is determined whether or not a scan instruction is input from the client PC 14, and the position of the original representing the original image set in the image reading unit 26 of the scanner device 12 is changed by the user. When a document is set in the image reading unit 26 of the scanner device 12 so that a part thereof overlaps the partial image S ₁ represented by the scanned partial image data A ₁ , and a scan instruction is input from the client PC 14, a step is performed. in 103, the scanning process is performed, partial read image data a ₂ represents a partial image S ₂ is generated. Similarly, the scanning process is performed in steps 104 and 105, partial read image data A ₃ represents a partial image S ₃ is generated.

In the next step 108, the partial read image data A ₁ , A ₂ , A ₃ are binarized according to the density to generate binary image data B ₁ , B ₂ , B ₃ , and binary data The image data B ₁ , B ₂ , B _{3 is} reduced by thinning out the data to generate reduced binary image data C ₁ , C ₂ , C _3. Thus, the inclination angle of the original image represented by the original set on the image reading unit 26 is detected for the partial images S ₁ , S ₂ , S ₃ represented by the reduced binary image data C ₁ , C ₂ , C ₃ .

Next, an inclination detection processing routine for realizing the processing of step 110 will be described with reference to FIG. First, in step 140, the reduced binary image data _{C 1, C 2, C 3} , performs contour extraction, to generate image data contour reduction binary _{D 1, D 2, D 3} . For example, the contour extraction process extracts any of the on pixel, the off pixel, and both the on pixel and the off pixel where the on pixel and the off pixel are adjacent to each other. In step 142, Hough transform processing is performed on the on-pixels of the contour reduced binary image data D ₁ , D ₂ , and D ₃ , and each of the contour reduced binary image data D ₁ , D ₂ , and D ₃ is processed. To generate Hough space data. The Hough transform process is a process of converting a straight line on the xy orthogonal coordinates into a coordinate value on the ρθ polar coordinate, and becomes a coordinate value of one point on the ρθ orthogonal coordinate as shown in FIG. In addition, one point on the xy orthogonal coordinates is converted into a line on the ρθ polar coordinate by the Hough transform process, and when the coordinate value on the polar coordinate on this line is expressed by the ρθ orthogonal coordinate, as shown in FIG. It can be represented by a curve. This curve can be expressed by the following equation.
ρ = x · cos θ + y · sin θ
Where θ is the angle with the x-axis, ρ is the distance from the origin of the xy orthogonal coordinates, and x and y are the coordinates on the xy orthogonal coordinates.

One point on the xy orthogonal coordinates representing the pixel position of the on-pixel of the contour reduced binary image data can be converted into a curve on the ρθ orthogonal coordinate by the above formula, and the contour reduced binary image data D ₁ , D ₂ and D ₃ , all the on-pixels are converted into a curve on the ρθ orthogonal coordinate by the above formula, and the frequency at which the coordinates (ρ, θ) are converted to a point on the curve is calculated. The obtained frequency becomes a histogram on the ρθ polar coordinate space, and this histogram is generated as the Hough space data for each of the contour reduced binary image data D ₁ , D ₂ , and D ₃ .

In the next step 144, a calculation projection process is performed in which the histogram of the Hough space data is projected onto the θ-axis to calculate the total frequency, and each of the reduced-contour binary image data D ₁ , D ₂ , and D ₃ described below is performed. The operation projection histogram data is generated. First, the diagonal length D (the square root of the sum of the square of the vertical length and the square of the horizontal length) of the image represented by the contour-reduced binary image data is obtained, and θ is π from 0 to π. The calculated histogram data, which is the frequency density, is obtained by the following equation.

  However, freq (θ, ρ) is the frequency of polar coordinates (θ, ρ) in the Hough space data.

In step 146, based on the calculated projection histogram data of each of the contour reduced binary image data D ₁ , D ₂ , and D ₃ , θ that maximizes hist (θ) is set as partial read image data A ₁ , A _2. , A ₃ are calculated as the inclination angles of the partial images S ₁ , S ₂ , S ₃ , and the inclination detection processing routine is terminated.

Then, in step 112 of the read image joining processing routine of FIG. 2, the erecting direction detection processing routine shown in FIG. 5 is executed. First, in step 160, character portion extraction processing is performed on the reduced binary image data C ₁ , C ₂ , and C ₃ . For example, a circumscribed rectangle is calculated based on the ON pixels of the reduced binary image data C ₁ , C ₂ , and C ₃ , a circumscribed rectangle that is estimated to be a character portion is specified from the aspect ratio of the circumscribed rectangle, and the circumscribed rectangle Is extracted as a character part. In step 162, character recognition processing is performed on the character portion extracted in step 160. Character data representing the character recognized from the character part and probability data representing the probability that the character displayed by the character part is recognized by the character recognition process are created, and the character direction of the character part is set to a predetermined angle, for example, 90 The character recognition processing is performed four times by rotating each time, character data and probability data are created for each of the four character directions, and this character recognition processing is executed for the plurality of character portions extracted in step 160.

In the next step 164, based on the plurality of probability data created in step 162, it is determined which character direction probability data is the highest. For example, all probability data is summed for each character direction, and the total value is compared to determine which character direction probability data is the highest in each of the reduced binary image data C ₁ , C ₂ , and C ₃ . The character direction having the highest probability data is calculated as the upright direction of each of the partial images S ₁ , S ₂ , S ₃ represented by the partial read image data A ₁ , A ₂ , A _3. The direction detection processing routine is terminated.

Then, in step 114 of the read image joining process routine of FIG. 2, the partial read image data A ₁ , A ₂ , A ₃ corrected in steps 110 and 112 are binarized according to the density to obtain binary image data. E ₁ , E ₂ , E ₃ are generated, and the thicknesses of the lines represented by the binary image data E ₁ , E ₂ , E ₃ due to the difference in density at the time of reading in steps 101, 103, 105 are made equal. Thin line binary image data F ₁ , F ₂ , and F ₃ are generated. Then, in step 116 extracts feature points by filtering the fine line binary image data F _1. For example, a pixel representing an intersection on a plurality of predetermined graphic patterns such as “+”, “⊥”, and “L” is searched as a feature point. Previously small rectangular area defined scans the thin line binary image data F ₁ above, by applying a filter corresponding to each of the plurality of graphic patterns in a small rectangular area of the figure pattern small rectangular region predetermined It is determined whether or not a contour is represented. At this time, based on the tilt angle calculated in step 110, the filter corresponding to each graphic pattern is changed to a filter corresponding to each graphic pattern tilted by the calculated tilt angle, and the changed filter is used. To make a decision. If it is determined that it corresponds to the contour of any graphic pattern, the pixel position information and the graphic pattern information located at the intersection on the contour are recorded in the feature point data. Since a plurality of feature points are extracted, a plurality of image position information and graphic pattern information are recorded in the feature point data. The thin line binary image data F ₂ and F ₃ are similarly filtered to extract feature points.

In the next step 117, based on the feature point data of the thin line binary image data F ₁ feature data and thin line binary image data F _2, 2 a feature that is common to the partial image S ₁ and the partial image S ₂ Search for a point. For example, a first common feature point having the same graphic pattern as the first feature point recorded in the feature point data of the fine line binary image data F ₁ is searched from the feature point data of the fine line binary image data F ₂ . .

When the first common feature point is retrieved, the same as the _second feature point recorded in the feature point data of the fine line binary image data F ₁ based on the feature point data of the fine line binary image data F 2 From the feature point of the graphic pattern and from the periphery of the pixel position moved from the pixel position of the first common feature point by the vector from the pixel position of the first feature point to the pixel position of the second feature point A second common feature point is searched. At this time, based on the erected direction calculated by the tilt direction and the step 112 calculated in step 110, the direction of the vector for erecting direction of the partial image S _1, the vector for erecting direction of the partial image S ₂ The second common feature point is searched from around the pixel position moved by this vector so that the directions are the same.

When the second common feature point is searched, the first common feature point and the second common feature point are recorded in the common feature point data, and the partial read image data A ₁ , to record the set of a _2. On the other hand, when the second common feature point is not searched, another first common feature point is searched from the feature point data of the thin line binary image data F ₂ , and if it is searched, the thin line binary image is searched. searching for a second common feature points from the feature point data of the data F _2.

The above search processing until the first common feature points and the second common feature point is retrieved, or, for all combinations of feature points recorded in the feature point data of the thin line binary image data F ₁ This is executed until the first common feature point and the second common feature point are searched.

In step 118, as in step 117, based on the feature point data of the thin line binary image data F _{1 and} the feature point data of the thin line binary image data F ₃ , the partial image S ₁ and the partial image S ₃ Two common feature points are searched, and when two common feature points are searched, the two feature points are recorded in the common feature point data, and the partial read image data A ₁ , to record the set of a _3. Also in the next step 119, as in step 117, based on the feature point data of the thin line binary image data F ₂ feature data and thin line binary image data F _3, the partial image S ₂ and the partial image Find the two feature points common to the S _3, the common two feature points is searched, and recording the two feature points to the common feature point data, partial read image data to the common feature point image data Record the set of A ₂ and A ₃ .

  In step 120, the set of partial images represented by the set of partially read image data recorded in the common feature point image data is joined so that the two feature points recorded in the common feature point data overlap, Original image data representing an image is generated. In step 122, the original image data is output to the client PC 14, and the read image joining processing routine is terminated.

  As described above, according to the scanner device according to the first embodiment, by detecting the inclination angle and the upright direction of each partial image with respect to the original image, the inclination angle and the positive direction of each partial image with respect to the original image are detected. Assuming a difference in vertical direction, there is no need to expand the range of searching for two feature points common to the set of partial images to be joined, and it is not necessary to increase the types of graphic patterns to be extracted as feature points Thus, it is possible to speed up the process of searching for two common feature points. In addition, it is possible to prevent an erroneous feature point from being searched as two common feature points due to a difference in inclination angle and erecting direction with respect to the original image of each partial image, so that the accuracy of the joining process can be improved.

  In the above-described first embodiment, the case where the inclination detection process is performed on the image that has been subjected to the contour extraction process from the binary image has been described as an example. May be subjected to differential filter processing to perform edge extraction processing, and the inclination may be detected from the edge extracted image.

  In addition, the case where tilt detection processing and erecting direction detection are performed on a reduced image has been described as an example, but tilt detection processing and erecting direction detection are performed on an image of the same size that is not subjected to reduction processing. You may go.

  Next, a second embodiment of the present invention will be described. The same parts as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

  In the second embodiment, the partially read image data is corrected based on the tilt angle detected by the tilt detection process described in the first embodiment and the erect direction detected by the erect direction detection process. The difference from the first embodiment is that feature points are extracted based on the corrected partial read image data and common feature points are searched. Note that the configuration of the second embodiment is the same as the configuration of the first embodiment, and a description regarding the configuration is omitted.

  Next, a read image joining process routine according to the second embodiment will be described with reference to FIG. Note that the same processes as those of the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

In the scanner device 12, first, in steps 100 and 101, a scan process of a part of the original image represented by the document is executed in the scanner device 12 to generate partial read image data A ₁ which is color image data. Thus, the original is set in the image reading unit 26 of the scanner device 12 so as to partially overlap the partial image S ₁ which is a part of the original image scanned in step 101, and the partial read image representing the partial image S ₂ data A ₂ is generated. Similarly, the scanning process is performed in steps 104 and 105, partial read image data A ₃ represents a partial image S ₃ is generated. In the next step 108, reduced binary image data C ₁ , C ₂ , C ₃ are generated, and in step 110, the inclination angle of the original image represented by the original set on the image reading unit 26 is determined as the partial images S ₁ , S 3. _2, to detect the S _3.

In step 200, the partial read image data A ₁ , A ₂ , A ₃ are subjected to rotation correction processing so that the inclination angles of the partial images S ₁ , S ₂ , S ₃ match. For example, rotation correction processing is performed on the partial read image data A ₁ , A ₂ , A _{3 so} that the partial images S ₁ , S ₂ , S ₃ are rotated in reverse by the respective inclination angles detected in step 146. Then, the inclination angles of the partial images S ₁ , S ₂ , S ₃ are set to zero. It should be noted that rotation correction processing is performed on each of the partial read image data A ₂ and A ₃ by a difference from the inclination angle of the partial image S ₁ detected in step 146, and the partial images S ₁ , S ₂ , S ₃ are processed. The inclination angles may be matched. Further, as a method of the rotation correction process, for example, a known method such as Affine conversion may be used, and thus detailed description of the rotation correction process is omitted.

In the next step 112, the erect directions of the partial images S ₁ , S ₂ , S ₃ are detected, and in step 202, the erect directions of the partial images S ₁ , S ₂ , S ₃ that have been rotationally corrected in step 200 are determined. Rotation correction processing is executed on the partial read image data A ₁ , A ₂ , A ₃ so as to match. At this time, the partial read image data A ₁ , A ₂ , and A ₃ are subjected to rotation correction processing so that the upright directions of the partial images S ₁ , S ₂ , and S ₃ are the upright directions of the original image. at best, also, to match the partial image S _2, S ₃ in the erecting direction of the partial image S _1, may be rotated correction process only partial images S _2, S _3.

In step 204, the partial read image data A ₁ , A ₂ , A ₃ corrected in steps 200, 202 are binarized according to the density to obtain binary image data E ₁ , E ₂ , E _3. , And thinning processing is performed to equalize the thicknesses of the lines represented by the binary image data E ₁ , E ₂ , and E ₃ , thereby generating thin line binary image data F ₁ , F ₂ , and F ₃ . In step 206, a predetermined small rectangular area is scanned on the thin line binary image data F ₁ , F ₂ , F ₃ and a filter corresponding to each of a plurality of graphic patterns is applied in the small rectangular area, thereby Feature points are extracted from each of the images S ₁ , S ₂ , S ₃ , and the feature point data in which the pixel position information and graphic pattern information of the feature points are recorded as the fine line binary image data F ₁ , F ₂ , F ₃ . Generate for each.

In the rotation correction process in step 200 described above, when only the partial read image data A ₂ and A ₃ are subjected to the rotation correction process so that the inclination angles of the partial images S ₁ , S ₂ , and S ₃ coincide with each other, rotating the plurality of figure pattern predetermined to correspond to the inclination angle of the partial image S _1, so as to correspond to the figure pattern is rotated, it is also changed filter used to extract feature points.

Then, in step 208, based on the feature point data of the thin line binary image data F _{1 and} the feature point data of the thin line binary image data F ₂ , _two features common to the partial image S ₁ and the partial image S ₂ are used. When a point is searched and two common feature points are searched, these two feature points are recorded in the common feature point data, and a set of partial read image data A ₁ and A ₂ is recorded in the common feature point image data. To do. Since the inclination angles and the erecting directions of the partial images S ₁ and S ₂ are corrected to coincide with each other, the vector direction from the first feature point to the second feature point in the partial image S ₁ is corrected. And the direction of the vector from the first common feature point to the second common feature point in the partial image S ₂ is the same.

Then, in step 210, based on the feature point data of the thin line binary image data F ₁ feature data and thin line binary image data F _3, 2 a feature that is common to the partial image S ₁ and the partial image S ₃ When a point is searched and two common feature points are searched, the two feature points are recorded in the common feature point data, and a set of partial read image data A ₁ and A ₃ is recorded in the common feature point image data. To do. Further, in the same manner as step 212, based on the feature point data of the thin line binary image data F ₂ feature data and thin line binary image data F _3, common to the partial image S ₂ and the partial image S ₃ 2 When two feature points are searched for, the two feature points are recorded in the common feature point data, and a set of partial read image data A ₂ and A ₃ is added to the common feature point image data. Record.

  In step 120, the set of partial images represented by the set of partial read image data recorded in the common feature point image data is joined so that the two feature points recorded in the common feature point data overlap, and the original image Is generated, and in step 122, the original image data is output to the client PC 14, and the read image joining processing routine is terminated.

  As described above, according to the scanner device according to the second embodiment, the rotation correction processing is performed on the plurality of partial read image data so that the inclination angle and the erecting direction with respect to the original image of each of the plurality of partial images coincide. By doing so, it is not necessary to expand the range for searching for a plurality of feature points common to the set of partial images to be joined, assuming the difference between the inclination angle and the erecting direction, and Since it is not necessary to increase the types of feature point graphic patterns in consideration of differences, it is possible to speed up the process of searching for a plurality of common feature points.

  Next, a third embodiment of the present invention will be described. The same parts as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

In the third embodiment, in addition to the inclination calculation process and the erecting direction calculation process described in the first embodiment, a process for calculating the arrangement positions of the partial images S ₁ , S ₂ , S ₃ with respect to the original image is performed. The point to be executed is different from the first embodiment. Note that the configuration of the third embodiment is the same as the configuration of the first embodiment, and a description regarding the configuration is omitted.

  Next, a read image joining process routine according to the third embodiment will be described with reference to FIG. Note that the same processes as those of the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

In the scanner device 12, first, in steps 100 and 101, a scanning process of a part of the original image represented by the document is executed in the scanner device 12 to generate partial read image data A _{1. In} steps 102 and 103, the partial image S is processed. partial read image data a ₂ represents a ₂ is generated. Similarly, the scanning process is performed in steps 104 and 105, partial read image data A ₃ represents a partial image S ₃ is generated. In the next step 108, reduced binary image data C ₁ , C ₂ , C ₃ are generated, and in step 110, the inclination angle of the original image represented by the original set on the image reading unit 26 is determined as the partial images S ₁ , S 3. _2, detects the S _3, in step 112, detects the erecting direction of the partial image _{S 1, S 2, S 3} .

In the next step 250, the arrangement position detection processing routine shown in FIG. 8 is executed. First, in step 270, feature points of a plurality of predetermined graphic patterns are extracted from the upper, lower, right and left regions of the partial image S ₁ based on the reduced binary image data C _1. Then, feature point data composed of graphic pattern information of feature points and pixel position information is created for each of the upper part, the lower part, the right part, and the left part. Similarly, for the partial images S ₂ and S ₃ , feature point data for the upper, lower, right, and left portions are generated. At this time, the filter for extracting the graphic pattern is corrected based on the inclination angles of the partial images S ₁ , S ₂ , S ₃ calculated in step 110, and the feature points are extracted from the partial images S ₁ , S ₂ , S _3. To extract.

In step 272, based on the feature point data of the partial image S ₁ feature data and the partial image S _2, executes the position detecting process of the partial image S ₁ and the partial image S _2. For example, based on the erect direction of each of the partial images S ₁ and S ₂ calculated in step 112, the upper feature point data viewed from the erect direction of the partial image S ₁ and the erect direction of the partial image S ₂ Based on the seen feature point data at the bottom, it is searched whether there are two common feature points. Similarly, saw partial image S ₁ of the erecting direction from the lower top of that seen from the erecting direction of the feature point data and the partial image S ₂ of viewed feature point data, the erecting direction of the partial image S ₁ feature point data of the left portion as viewed from the erecting direction of the feature point data and the partial image S ₂ of the right portion, and a part left of the feature point data as seen from the upright direction of the image S ₁ and the partial positive image S ₂ Two feature points common to the partial image S ₁ and the partial image S ₂ are searched based on the feature point data on the right side as viewed from the vertical direction. The arrangement position information of the partial image represented by the set of partial read image data having one feature point and the approximate common point information including the pixel position information and the graphic pattern information of the two feature points are created.

In the next step 274, similarly to step 272, the arrangement position detection processing of the partial image S ₁ and the partial image S ₃ is performed based on the feature point data of the partial image S _{1 and} the feature point data of the partial image S _3. When two feature points common to the partial image S ₁ and the partial image S ₃ are searched, arrangement position information and approximate common point information are generated. In step 276, the characteristic points of the partial image S ₂ are generated. based on the data and the partial feature point data of the image S _3, perform the position detecting process of the partial image S ₂ and the partial image S _3, 2 a feature that is common to the partial image S ₂ and the partial image S ₃ When a point is searched, arrangement position information and general common point information are generated, and the arrangement position detection processing routine is terminated.

Then, in step 114 of the read image joining process routine of FIG. 7, the binarization process and the thinning process are executed on the partial read image data A ₁ , A ₂ , A ₃ to obtain the thin line binary image data F ₁ , F ₂ and F ₃ are generated. In step 251, feature points are obtained from the partial images S ₁ , S ₂ and S ₃ based on the approximate common point information and the thin line binary image data F ₁ , F ₂ and F _3. Extraction is performed to generate feature point data for each of the partial images S ₁ , S ₂ , S ₃ . At this time, only feature points having the same graphic pattern information as the graphic pattern information of the approximate common point information from the periphery of the pixel position in the thin line binary image data represented by the pixel position information of the approximate common point information generated in step 250 above. To extract.

  In the next step 252, 1 is set as an initial value to a number n indicating a set of thin-line binary image data representing a set of partial images having two common feature points. In the next step 254, the n-th thin line is set. Two feature points common to the set of partial images represented by the set of binary image data are searched, the two searched feature points are recorded in the common feature point data, and the partially read image having the two common feature points The data set is recorded in the common feature point image data.

  In the next step 256, it is determined whether or not there is an n + 1th thin line binary image data set. If the determination is affirmative, the value of n is incremented in step 258 and the process returns to step 254. When the process of searching for the common feature point in step 254 is executed for the set of thin line binary image data, the determination in step 256 is denied, and in step 120, the partially read image data recorded in the common feature point image data is determined. The set of partial images represented by the set is joined so that the two feature points recorded in the common feature point data are overlapped to generate original image data representing the original image. In step 122, the original image data is sent to the client PC. Output, and the read image joining processing routine ends.

  As described above, according to the scanner device according to the third embodiment, a plurality of feature points common to a set of partial images to be joined are searched by detecting the arrangement positions of the partial images with respect to the original image. Since the range to be performed can be limited, it is possible to speed up the process of searching for a plurality of common feature points.

  In the above embodiment, the feature points are extracted from each of the thin line binary image data, and two common feature points are searched from the extracted feature points. Two thin line binary image data to be searched for feature points are set as a reference side thin line binary image data and a search side thin line binary image data, and a plurality of features are obtained from the reference side thin line binary image data. Two common feature points may be searched by extracting a point and determining whether the same feature point as the extracted feature point can be extracted from the thin-line binary image data on the search side.

1 is a schematic diagram showing a configuration of an image processing system according to a first embodiment of the present invention. It is a flowchart which shows the read image joining process routine which concerns on the 1st Embodiment of this invention. It is a flowchart which shows the inclination detection process routine which concerns on the 1st Embodiment of this invention. It is an image figure explaining the content of the Hough conversion process in the inclination detection process routine which concerns on the 1st Embodiment of this invention. It is a flowchart which shows the erecting direction detection process routine which concerns on the 1st Embodiment of this invention. It is a flowchart which shows the read image joining process routine which concerns on the 2nd Embodiment of this invention. It is a flowchart which shows the read image joining process routine which concerns on the 3rd Embodiment of this invention. It is a flowchart which shows the arrangement position detection processing routine based on the 3rd Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Image processing system 12 Scanner apparatus 14 Client PC
20 ROM
22 CPU
24 RAM
26 Image reading unit 28 HDD
A ₁ , A ₂ , A ₃ partial read image data S ₁ , S ₂ , S ₃ partial image

Claims (6)

Detecting means for detecting information indicating a state of each of the partial images represented by each of the plurality of partial read image data obtained by dividing the original image into a plurality of times so as to partially overlap the original image; ,
Search means for searching a set of partial images having a plurality of common feature points from the plurality of partial images based on the detected information and a plurality of feature points in each of the partial images;
A joining generation unit that joins the searched partial image sets so that the plurality of common feature points overlap, and generates original image data representing the original image;
An image processing apparatus.   The detection means includes, as information indicating a state with respect to the original image, an inclination angle of each of the partial images with reference to a predetermined direction of the original image, an upright direction of each of the partial images with respect to the original image, and the original image The image processing apparatus according to claim 1, wherein at least one position of each of the partial images with respect to the image is detected.   The search means extracts a plurality of feature points in each of the partial images by processing the plurality of partial read image data using a filter corresponding to the detected information, and the detected information and the The image processing apparatus according to claim 1, wherein a set of partial images having a plurality of common feature points is searched from the plurality of partial images based on a plurality of feature points in each of the extracted partial images.   The search means corrects the plurality of partial read image data based on the detected information so that a state of each of the plurality of partial images with respect to the original image matches, and then corrects the plurality of partial scans corrected. Based on the image data, a plurality of feature points in each of the partial images is extracted, and based on the plurality of feature points in each of the extracted partial images, a plurality of common feature points from the plurality of partial images are extracted. The image processing apparatus according to claim 1, wherein a set of partial images is searched. The detection means has at least one of an inclination angle of each of the partial images with reference to a predetermined direction of the original image, and an erecting direction of each of the partial images with respect to the original image as information indicating a state with respect to the original image. Detect
The search means includes the plurality of the plural images so that an inclination angle of each of the partial images with reference to a predetermined direction of the detected original image matches at least one of an erect direction of each of the partial images with respect to the original image. A plurality of feature points in each of the partial images is extracted based on the rotation-corrected partial read image data, and a plurality of feature points in each of the extracted partial images are extracted. The image processing apparatus according to claim 1, wherein a set of partial images having a plurality of common feature points is searched from the plurality of partial images based on the feature points. Detecting information indicating a state of each of the partial images represented by each of the plurality of partial read image data obtained by dividing the original image into a plurality of times so as to partially overlap,
Based on the detected information and a plurality of feature points in each of the partial images, a set of partial images having a plurality of common feature points is searched from the plurality of partial images,
An image processing method comprising: combining the searched partial images so that the plurality of common feature points overlap each other to generate original image data representing the original image.

Images