JP5589825B2 - Image processing apparatus, image processing method, and image processing program - Google Patents

Description

  The present invention relates to an image processing apparatus and the like.

  In recent years, paper documents have been digitized using an OCR (Optical Character Reader). By digitizing paper documents, users can save the storage space for documents. Further, by digitizing a paper document, it is possible to easily search for the document or reuse the contents of the document.

  Here, in order for the OCR to digitize a paper document and create an electronic document corresponding to the layout of the paper document, it is necessary to cut out characters, words, lines, character blocks, and the like from the document image.

  In the prior art, there is a method in which a histogram obtained by projecting black pixels of a document image is calculated, and a line is cut out from the document image by comparing the histogram with each threshold value. 12-14 is a figure for demonstrating a prior art. A document image 1a in FIG. 12 is an image obtained by digitizing a paper document. When the histogram is calculated by projecting the black pixels of the document image 1a, the histogram 1b of FIG. 13 is obtained.

  In the prior art, after calculating the histogram 1b, smoothing is performed by deleting a convex portion having a width T1 or less and a concave portion having a width T2 or less. When smoothing is performed on the histogram 1b, the histogram 1c of FIG. 14 is obtained. In the conventional technique, the absolute value of the difference between adjacent histogram values is compared with the threshold T3, and a position where the absolute value exceeds the threshold T3 is determined as a row. In the example illustrated in FIG. 14, the line between the position a and the position b is determined as a row.

  In the above conventional technique, many parameters are used when a word, line, or document block is cut out from a document image. In the prior art shown in FIGS. 12 to 14, parameters T1 to T3 are required only by specifying a row. When a word or character block is cut out, the user prepares more parameters.

Akiyama Teruo, Masuda Isao, “A Method for Extracting Constituent Elements of Paper Not Using Format Designation Information”, IEICE Journal J66-D, No. 1

  By the way, an image generally has a different observed structure depending on a scale. Hereinafter, this scale is expressed as a scale. The scale corresponds to the degree of distortion when the image is blurred, and when scaled by a Gaussian filter using a Gaussian function, it is equal to the variance of the Gaussian function. Furthermore, an image that has been blurred at a certain scale value is referred to as a scale image at that scale. Moreover, the most suitable scale for observing a certain structure is described as the inherent scale of the structure. In a document image, this inherent scale may be obtained globally. FIG. 15 is a diagram illustrating an example of document images having different structures that are observed according to the scale. As shown in FIG. 15, when the scale is 2a, the observed structure is a single letter “F”, and when the scale is 2b, the observed structure is a set of letters “E”, and the scale is 2c. Then, the observed structure becomes a set of squares.

  When cutting out characters, texture elements, and the like from a document image as shown in FIG. 15, it is necessary for the user to appropriately adjust the scale value with reference to the document image, and this imposes a burden on the user. It was big. Note that the conventional technique for determining characters, lines, and the like using the histogram described above sets parameters for cutting out characters, lines, etc. by the user himself. For this reason, in the prior art, when characters and texture elements are cut out from the document image shown in FIG. 15, the user needs to set each parameter so that the characters and texture elements can be recognized at a desired scale. The burden on users cannot be reduced.

  The disclosed technique has been made in view of the above, and is capable of automatically deriving a global eigenscale and determining a region of an image element corresponding to a structure observed in the global eigenscale. It is an object to provide an image processing apparatus, an image processing method, and an image processing program that can be automatically obtained.

  The disclosed image processing apparatus includes a scale image generation unit that generates a plurality of scale images having different degrees of blur from the image. In addition, the disclosed image processing apparatus includes a blob image generation unit that generates, as a blob image, a set of pixels having the same density in the scale image in each scale image. Further, the disclosed image processing apparatus has a unique scale for calculating a degree of blurring in a range in which a ratio of a change in the area or number of blob images to a change in a value indicating the degree of blurring of an image falls within a predetermined threshold. It has a calculation part. The disclosed image processing apparatus further includes an output unit that outputs a blob image area extracted from the scale image corresponding to the degree of blur calculated by the inherent scale calculation unit as an area for character recognition. Is a requirement.

  According to the disclosed image processing apparatus, a global eigenscale can be automatically derived, and a region of an image element corresponding to a structure observed in the global eigenscale can be automatically obtained. .

FIG. 1 is a diagram illustrating a configuration of an image processing apparatus. FIG. 2a shows a document image. FIG. 2 b is a diagram illustrating each scale image of the document image with the inherent scale values t _{0 to} t ₃ and each blob image obtained from the scale image. FIG. 3 is a diagram showing the relationship between the number of blobs and the scale value of the image of FIG. 2a. FIG. 4 is a diagram showing a calculation result of the first derivative of the graph of FIG. FIG. 5 is a diagram (1) illustrating an example of an image element region extracted from each scale image. FIG. 6 is a diagram (2) illustrating an example of an image element region extracted from each scale image. FIG. 7 is a diagram (3) illustrating an example of an image element region extracted from each scale image. FIG. 8 is a diagram (4) illustrating an example of an image element region extracted from each scale image. FIG. 9 is a flowchart illustrating a processing procedure of processing for calculating the unique scale value. FIG. 10 is a flowchart showing a processing procedure of processing for cutting out an image of the image element area. FIG. 11 is a diagram illustrating a hardware configuration of a computer constituting the image processing apparatus according to the embodiment. FIG. 12 is a diagram (1) for explaining the prior art. FIG. 13 is a diagram (2) for explaining the prior art. FIG. 14 is a diagram (3) for explaining the prior art. FIG. 15 is a diagram illustrating an example of an image having a hierarchical structure with different structures observed according to the scale. FIG. 16 is a diagram for explaining processing of the image element region determination unit.

  Embodiments of an image processing apparatus, an image processing method, and an image processing program disclosed in the present application will be described below in detail with reference to the drawings. Note that the present invention is not limited to the embodiments.

  The configuration of the image processing apparatus according to the present embodiment will be described. FIG. 1 is a diagram illustrating a configuration of an image processing apparatus. As illustrated in FIG. 1, the image processing apparatus 100 includes an input unit 110, an image reading unit 120, a display unit 130, a scale calculation processing unit 140, and an image element cutout unit 150.

  The input unit 110 corresponds to an input device such as a keyboard or a mouse. The image reading unit 120 reads an image from a paper medium or the like, and outputs the read image data to the scale calculation processing unit 140 and the image element clipping unit 150. For example, the image reading unit 120 corresponds to a scanner. In this embodiment, the handled image is a gray image or a monochrome binary image. Note that when a color image is input, the image processing apparatus 100 may perform brightness conversion and convert the color image into a gray image. The display unit 130 corresponds to a display device such as a display.

  The scale calculation processing unit 140 is a processing unit that calculates a global scale unique to image data having a hierarchical structure. The scale calculation processing unit 140 includes a scale image generation unit 141, a blob image generation unit 142, and a unique scale calculation unit 143.

  The scale image generation unit 141 is a processing unit that generates a plurality of scale images by applying a Gaussian filter with scale values changed in multiple stages to image data. The scale image generation unit 141 outputs a plurality of scale images to the blob image generation unit 142 in association with the corresponding scale values.

Here, the Gaussian filter used by the scale image generation unit 141 is a filter defined using a Gaussian function. This Gaussian filter is used when blurring an image. Equation (1) represents a two-dimensional Gaussian function g. In formula (1), t represents a scale value. This scale value corresponds to the degree of blur. X and y indicate coordinates on the image.

When the original image is represented by I, the relationship between the original image I and the image B after being blurred by a Gaussian function is expressed by the following equations (2) and (3). In equation (2), (x, y) is the coordinates of the pixel of interest.

Here, the equation (2) holds in an ideal case where the images are continuous and the size is infinite. Actual images are discrete and have a finite size. For this reason, the relationship between the original image I and the image B is represented by the following formulas (4) and (5). As in Expression (4), the relationship between the original image I and the image B is represented by a sum instead of integration.

In equations (4) and (5), a finite size matrix is used instead of a Gaussian function as a Gaussian filter. The height and width of the Gaussian filter are 2s + 1. s is an integer. In equation (4), (i, j) is the coordinates of the pixel of interest. The Gaussian filter size 2s + 1 s is generally designed as a constant multiple of t. For example, the Gaussian filter when s = t, t = 1 and the center of the Gaussian filter is (m, n) = (0, 0) is expressed by Expression (6).

  The blob image generation unit 142 is a processing unit that generates a blob image from each scale image. The blob image generation unit 142 binarizes the scale image and performs connected component analysis, thereby setting one connected component as one blob. This blob can be said to be a collection of pixels darker or brighter than the background in the scale image. For each extracted blob, the blob image generation unit 142 sets an image to which a value different from the background is assigned, for example, if the background area is 1 and the blob area is 0, as the blob image.

  For example, the blob image generation unit 142 compares each pixel value of the scale value with a threshold value, and generates a blob image by binarizing the scale image based on the comparison result. For example, the blob image generation unit 142 binarizes the scale image by setting a pixel value equal to or greater than the threshold value to 1 and setting a pixel value smaller than the threshold value to 0 for each pixel of the scale image. The blob image generation unit 142 generates a blob image for each scale image.

FIG. 2a shows a document image. FIG. 2B is a diagram illustrating each scale image obtained from the document image with the scale values t _{0 to} t ₃ and each blob image obtained from the scale image. The magnitude relationship between the scale values is t ₀ <t ₁ <t ₂ <t ₃ . As shown in FIG. 2b, blob image generation unit 142, a halo or has been scaled image 10b by the scale value _{t 0} binarized to generate a blob image 10c. When the scale value t ₀ and 0, blob image generating unit 142, may be used as the document image in Figure 2a.

As shown in the second row from the top in FIG. 2b, blob image generation unit 142, a halo or has been scaled image 10d by the scale factor t ₁ binarized to generate a blob image 10e. As shown in the third row from the top in FIG. 2b, blob image generation unit 142, a halo or has been scaled image 10f by the scale value t ₂ binarized to generate a blob image 10g. As shown in the fourth stage from the top in FIG. 2b, blob image generation unit 142, a halo or has been scaled image 10h by the scale value t ₃ is binarized to generate a blob image 10i. As shown in FIG. 2b, as the scale value increases, the number of blobs decreases and the average value of the blob area increases.

  The blob image generation unit 142 generates a blob image for each scale image, then generates overall blob information for each scale image, and outputs the generated overall blob information to the eigenscale calculation unit 143.

  Here, the entire blob information includes a scale value, a scale image, a blob image, the number of blobs, and individual blob information. This individual blob information exists for each blob included in the same blob image. The individual blob information includes an array of blob areas, coordinates of vertices of circumscribed rectangles of the blob, barycentric coordinates of the blob, and blob pixel coordinates.

For example, the overall blob information corresponding to the fourth row from the top in FIG. 2a will be described. The overall blob information includes the scale value “t ₃ ”, the scale image 10 h, the blob image 10 i, the blob number “3”, the blob 1 a, A plurality of individual blob information respectively corresponding to 1b and 1c is included.

  Returning to the description of FIG. The unique scale calculation unit 143 is a processing unit that calculates a plurality of types of unique scale values based on the entire blob information. Each unique scale value is a value used when an image element is cut out from image data. For example, the unique scale value is used when a region of characters, words, lines, and character blocks included in a document image is cut out. The unique scale calculation unit 143 outputs each unique scale value, the scale image corresponding to the unique scale value, the entire blog information corresponding to the unique scale value, and the original document image to the image element cutout unit 150.

  Here, the relationship between the number of blobs included in the blob information and the scale value will be described. FIG. 3 is a diagram showing the relationship between the number of blobs and the scale value of the image of FIG. 2a. The horizontal axis in FIG. 3 is an axis corresponding to the scale value t, and the vertical axis is an axis corresponding to the logarithm ln f (t) of the blob number when the blob number is regarded as a function of the scale value. For example, when the relationship between the number of blobs and the scale value is obtained from an image having a hierarchical structure such as a document image, a steady state occurs in which the number of blobs does not change even if the scale value is changed. In the example shown in FIG. 3, the portions 20a, 20b, 20c, and 20d are in a steady state. This steady state is called a plateau. It should be noted that the number of blobs rapidly changes between different steady states, and this state is referred to as state transition.

The inherent scale calculation unit 143 calculates a scale value corresponding to the starting point of the steady state as a fixed scale value. In the example shown in FIG. 3, the scale values corresponding to the start point of the steady-state 20a becomes t _0. Furthermore, the scale value corresponding to the start point of the steady-state 2b is t _1. Furthermore, the scale value corresponding to the start point of the steady-state 2c is t _2. Furthermore, the scale value corresponding to the start point of the steady-state 20d is t _3. In this case, the inherent scale calculation unit 143 calculates the scale values t ₀ , t ₁ , t ₂ , and t ₃ as inherent scales.

The processing of the inherent scale calculation unit 143 will be specifically described. First, the eigenscale calculation unit 143 numerically calculates the first derivative df (t) / dt and the second derivative d ² f (t) / dt with respect to a graph indicating the relationship between the logarithm of the blob number and the scale value. To calculate. FIG. 4 is a diagram showing a calculation result of the first derivative of the graph of FIG.

After calculating the first derivative and the second derivative, the inherent scale calculator 143 detects all the scale values t _m at which the first derivative is minimized. The point at which the first derivative is minimized is equivalent to the value of the second derivative being zero. In the example shown in FIG. 4, the scale values at which the first derivative is minimized are t _M0 , t _M1 , and t _M2 .

The eigenscale calculation unit 143 obtains a scale value at which the first derivative is minimized, and then gradually increases the scale value to calculate a scale value at which the absolute value of the first derivative becomes smaller than a predetermined threshold value for the first time. To do. This scale value becomes the scale value corresponding to the starting point of the steady state. In the example shown in FIG. 4, the scale values at which the absolute value of the first derivative becomes smaller than the predetermined threshold for the first time are t ₁ , t ₂ , and t ₃ .

Also, unique scale calculator 143 detects the initial value t ₀ of the scale values. The scale value t ₀ corresponds to the beginning of the steady state 20a in FIG. As t 0 _{= 0,} may be the brightness image of the document image not umbrella halo. The inherent scale calculation unit 143 outputs the scale values t ₀ , t ₁ , t ₂ , and t ₃ as inherent scale values to the image element cutout unit 150. When the number of scale values whose absolute value of the first derivative becomes smaller than a predetermined threshold value for the first time is M, M + 1 scale values including the initial values of the scale values are detected as unique scale values. .

  Returning to the description of FIG. The image element cutout unit 150 is a processing unit that cuts out an image of the image element region from the image data using the unique scale value. For example, the image in the image element area corresponds to a character, a word, a line, and a character block. The image element cutout unit 150 includes an image binarization unit 151, an image element region determination unit 152, and a cutout unit 153. The image element cutout unit 150 is an example of an output unit.

  The image binarization unit 151 is a processing unit that binarizes pixels included in the image data into white pixels and black pixels when image data is acquired from the image reading unit 120. The image binarization unit 151 outputs image data obtained by binarizing the document image to the image element region determination unit 152.

  The image element region determination unit 152 is a processing unit that determines the region of the image element included in the image data based on the image data, the unique scale value, and the entire blob information. For example, the image elements included in the document image data correspond to characters, words, lines, and character blocks.

  Specifically, the processing of the image element region determination unit 152 will be described. The image element region determination unit 152 performs a labeling process on the binarized document image and extracts a circumscribed rectangle of each connected component. The image element region determination unit 152 compares the binarized document image with the circumscribed rectangle of each blob based on the entire blob information, and determines a connected component included in the circumscribed rectangle of the blob. Then, the image element region determination unit 152 extracts a circumscribed rectangle of the determined entire connected component as an image element region on the corresponding inherent scale. The image element area determination unit 152 outputs the extracted coordinates of the circumscribed rectangle to the clipping unit 153 as the coordinates of the image element area.

  FIG. 16 is a diagram for explaining processing of the image element region determination unit. FIG. 16 2a corresponds to a blob of a certain inherent scale. As shown in 2b, the image element region determination unit 152 compares the binarized document image with the circumscribed rectangle of the blob, and the connected component included in the circumscribed rectangle of the blob is “Book”. The image element region determination unit 152 extracts the circumscribed rectangle 2c of the entire determined connected component as an image element region on the corresponding inherent scale.

  Note that the image element region determination unit 152 excludes characters whose center of gravity is not included in the circumscribed rectangle of the blob. For example, in 2b of FIG. 16, the center of gravity of “B” of “Book” is not included in the circumscribed rectangle of the blob, and the centers of gravity of “o”, “o”, and “k” are included in the circumscribed rectangle of the blob. The circumscribed rectangle of the connected component is the circumscribed rectangle of “ok”. Note that the image element region determination unit 152 may obtain the center of the character from the circumscribed rectangle of the character and use the center of the character instead of the center of gravity.

5 to 8 are diagrams illustrating an example of image element regions determined from each blob image. FIG. 5 shows an image element area on the document image determined from the blob image corresponding to the unique scale value t ₁ . The image element area shown in FIG. 5 corresponds to a word area. 6 was determined from the blob image corresponding to the specific scale value t _2, shows an image element regions on the document image. The image area element area shown in FIG. 6 corresponds to a line area.

7 was determined from the blob image corresponding to the specific scale value t _3, shows an image element regions on the document image. The image region element region shown in FIG. 7 corresponds to a character block region. Figure 8 were determined from blob image corresponding to the specific scale value t _4, shows an image element regions on the document image.

  The cutout unit 153 is a processing unit that cuts out the image element region image from the original document image or the binary image of the document image based on the coordinates of the image element region. The image of the image element area corresponds to, for example, a character, a word, a line, and a character block. The cutout unit 153 outputs the cut image element region image to the display unit 130. The image element region cut out by the cutout unit 153 corresponds to the circumscribed rectangle of the entire connected component of the binary image of the original document image in which the center of gravity is included in the region of the blob image generated by each unique scale value. Note that the cutout unit 153 may output the image element region image data to a device that performs character recognition and the like.

  Incidentally, when the image to be cut out of the image element area is an image having a hierarchical structure as shown in FIG. 15, the image element cutout unit 150 includes the image element area 2a, the plurality of image element areas 2b, and the like. It will be cut out. The image of the image element area cut out by the image element cutout unit 150 is an area cut out according to the purpose such as character recognition.

  Next, a processing procedure of processing in which the image processing apparatus 100 calculates the unique scale value will be described. FIG. 9 is a flowchart illustrating a processing procedure of processing for calculating the unique scale value. The process illustrated in FIG. 9 is executed, for example, when the image reading unit 120 reads image data. As shown in FIG. 9, the image processing apparatus 100 generates a plurality of scale images by applying Gaussian filters having different scale values t to the image data (step S101).

  The image processing apparatus 100 extracts the entire blob image information from each scale image (step S102). The image processing apparatus 100 calculates a unique scale value based on the blob image state change with respect to the scale change (step S103). Then, the image processing apparatus 100 outputs a unique scale value (step S104).

  Next, a processing procedure of processing in which the image processing apparatus 100 cuts out an image of the image element area will be described. FIG. 10 is a flowchart showing a processing procedure of processing for cutting out an image of the image element area. The process illustrated in FIG. 10 is executed, for example, when the image element cutout unit 150 acquires the unique scale value from the scale calculation processing unit 140.

  As shown in FIG. 10, the image processing apparatus 100 acquires a unique scale value (step S201), and selects an unselected unique scale value (step S202). The image processing apparatus 100 calculates an image element region at the selected unique scale value (step S203).

  The image processing apparatus 100 cuts out the image element region image from the original document image or the binary image of the original document image (step S204), and determines whether all the unique scale values have been selected (step S205). . If all the unique scale values have not been selected (step S205, No), the image processing apparatus 100 proceeds to step S202. On the other hand, when all the unique scale values are selected (step S205, Yes), the image processing apparatus 100 outputs the cut image (step S206).

  Next, effects of the image processing apparatus 100 according to the present embodiment will be described. The image processing apparatus 100 generates a plurality of scale images having different scale values from the image data, and generates a blob image from each scale image. Then, based on the relationship between the number of blob images and the scale value, the image processing apparatus 100 calculates a scale value at which the rate of change in the number of blob images falls within a predetermined threshold value as a unique scale value. The image processing apparatus 100 sets a circumscribed rectangle of the entire connected component of the binary image of the document image in which the centroid is included in the blob of the blob image generated based on the unique scale value as an image element region in the unique scale, and the corresponding document image. Or a binary image area of the document image is output as an image element on the inherent scale. For this reason, even if the image is an image having a hierarchical structure, it is possible to automatically obtain a unique scale value to which an image element region should be cut out in each hierarchical structure.

  For example, if the image processing apparatus 100 according to the present embodiment is used, an area for character recognition can be automatically output even for an image having a hierarchical structure as shown in FIG. The image processing apparatus 100 can automatically obtain the area 2a for character recognition and the plurality of areas 2b from the image shown in FIG.

  Further, if the image processing apparatus 100 according to the present embodiment is used, it is possible to automatically output areas of characters, words, lines, and character blocks from an image having a hierarchical structure such as a document image. Among these, information on areas such as words, lines, and character blocks can be used when the image layout is restored to electronic data. The information on the word area can be used not only for restoring the layout, but also when checking the spelling of the word against the character recognition result for the area.

  Incidentally, the eigenscale calculation unit 143 calculates the eigenscale value based on the relationship between the number of blobs and the scale value, but is not limited to this. For example, the unique scale calculation unit 143 may calculate the unique scale value based on the relationship between the average area of the blob and the scale value. That is, the eigenscale calculation unit 143 may detect a steady state where the ratio of the average area of the blob does not change with respect to the change in the scale value, and calculate the scale value corresponding to the steady state as the eigenscale value. .

  In addition, the unique scale calculation unit 143 may calculate the unique scale value based on the relationship between the number of maximum points of the scale image and the scale value. For example, the eigenscale calculation unit 143 detects a steady state in which the number of local maximum points of the blob image does not change with respect to the change in the scale value, and calculates the scale value corresponding to the steady state as the eigenscale value. Good.

  Note that the scale calculation processing unit 140 and the image element cutout unit 150 illustrated in FIG. 1 correspond to an integrated device such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). Alternatively, the processing units 140 and 150 correspond to electronic circuits such as a CPU (Central Processing Unit) and an MPU (Micro Processing Unit).

  The image processing apparatus 100 can also be realized by mounting each function of the image processing apparatus 100 on an information processing apparatus such as a known personal computer, workstation, mobile phone, PHS terminal, mobile communication terminal, or PDA. .

  FIG. 11 is a diagram illustrating a hardware configuration of a computer constituting the image processing apparatus according to the embodiment. As shown in FIG. 11, the computer 30 includes a CPU 31 that executes various arithmetic processes, an input device 32 that receives input of data from a user, and a monitor 33. The computer 30 includes a medium reading device 34 that reads a program and the like from a storage medium, and a network interface device 35 that exchanges data with other computers via a network. The computer 30 also includes a scanner 36 that reads an image, a RAM 37 that temporarily stores various information, and a hard disk device 38. Each device 31 to 38 is connected to a bus 39.

  The hard disk device 38 stores a unique scale calculation program 38a and an image element cutout program 38b. The inherent scale calculation program 38 a has the same function as the scale calculation processing unit 140. The image element cutout unit 38b has the same function as the image element cutout unit 150.

  The CPU 31 reads the inherent scale calculation program 38 a and the image element cutout program 38 b from the hard disk device 38 and expands them in the RAM 37. The inherent scale calculation program 38a functions as an inherent scale calculation process 37a. The image element cutout program 38b functions as an image element cutout process 37b.

  The unique scale calculation process 37a acquires image data from the scanner 34, and calculates a unique scale value based on the image data. The image element cutout process 37b cuts out an image of the image element area from the image data using the unique scale value.

  The inherent scale calculation program 38a and the image element cutout program 38b are not necessarily stored in the hard disk device 38. For example, the computer 30 may read and execute a program stored in a storage medium such as a CD-ROM. Alternatively, the program may be stored in a public line, the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), or the like, and the computer 30 may read and execute the program.

  The following supplementary notes are further disclosed with respect to the embodiments including the above examples.

(Supplementary Note 1) a scale image generation unit that generates a plurality of scale images having different degrees of blurring from an image;
In each scale image, a blob image generation unit that generates a blob image including a plurality of blobs by comparing each pixel value in the scale image with a predetermined threshold value and binarizing the scale image; Based on the relationship between the value indicating the degree of blur and the area or number of each blob, the ratio of the change in the average area or number of blobs to the change in the value indicating the degree of blurring of the image is less than a predetermined threshold. An eigenscale calculator that calculates the value of the extent of the range;
An image processing apparatus comprising:

(Supplementary Note 2) The image is a document image, the document image is compared with a blob area included in the blob image, and a circumscribed rectangular area of a character or a character string included in the blob area is converted into a unique scale. The image processing apparatus according to appendix 1, further comprising an output unit that outputs an image element region in

(Supplementary Note 3) The output unit compares the centroid of the character included in the blob area or the centroid of each character of the character string with the blob area, and the character or character string including the centroid in the blob area The image processing apparatus according to appendix 2, wherein the circumscribed rectangular area is output as an image element area on a specific scale.

(Supplementary Note 4) The output unit compares the blob area of each blob image generated from the scale images having different scrambles with the document image, so that the character area, word area, line 4. The image processing apparatus according to appendix 2 or 3, wherein an area and a character block area are output.

(Supplementary Note 5) An image processing method executed by a computer,
Generate multiple scale images with different degrees of blur from the image,
In each scale image, each pixel value in the scale image is compared with a predetermined threshold value, and the scale image is binarized to generate a blob image including a plurality of blobs.
Based on the relationship between the value indicating the degree of blurring of the image and the area or number of each blob, the ratio of the change in the average area or number of blobs to the change in the value indicating the degree of blurring of the image is a predetermined value. An image processing method comprising calculating a degree of blurring in a range that is less than a threshold value.

(Supplementary Note 6) The image is a document image, the document image is compared with a blob area included in the blob image, and a circumscribed rectangular area of a character or a character string included in the blob area is converted into a unique scale. 6. The image processing method according to appendix 5, wherein the image element area is output as an image element area.

(Supplementary note 7) The center of the character included in the blob region or the center of each character of the character string is compared with the region of the blob, and the circumscribed rectangle region of the character or character string including the center of gravity in the blob region Is output as an image element area on a specific scale.

(Supplementary note 8) By comparing the blob area of each blob image generated from the scale images having different frauds with the document image, a character area, a word area, a line area, a character block The image processing method according to appendix 6 or 7, characterized in that:

(Appendix 9)
Generate multiple scale images with different degrees of blur from the image,
In each scale image, each pixel value in the scale image is compared with a predetermined threshold value, and the scale image is binarized to generate a blob image including a plurality of blobs.
Based on the relationship between the value indicating the degree of blurring of the image and the area or number of each blob, the ratio of the change in the average area or number of blobs to the change in the value indicating the degree of blurring of the image is a predetermined value. An image processing program for executing a process of calculating a degree of blur in a range that is less than a threshold.

(Supplementary note 10) The image is a document image, the document image is compared with a blob area included in the blob image, and a circumscribed rectangular area of a character or a character string included in the blob area is converted into a unique scale. The image processing program according to appendix 9, wherein the image processing area is output as an image element area.

(Supplementary Note 11) The circumscribed rectangle region of the character or character string whose center of gravity is included in the blob region by comparing the center of gravity of the character included in the region of the blob or the center of gravity of each character of the character string with the region of the blob Is output as an image element area on a specific scale.

(Additional remark 12) By comparing the blob area of each blob image generated from the scale image having a different degree of blurring with the document image, a character area, a word area, a line area, a character block The image processing program according to appendix 10 or 11, wherein the image processing program is output.

DESCRIPTION OF SYMBOLS 100 Image processing apparatus 110 Input part 120 Image reading part 130 Output part 140 Scale calculation process part 150 Image element cutout part

Claims (6)

A scale image generation unit for generating a plurality of scale images having different degrees of blur from the image;
In each scale image, a blob image generation unit that generates a blob image including a plurality of blobs by comparing each pixel value in the scale image with a predetermined threshold value and binarizing the scale image; Based on the relationship between the value indicating the degree of blur and the area or number of each blob, the ratio of the change in the average area or number of blobs to the change in the value indicating the degree of blurring of the image is less than a predetermined threshold. An image processing apparatus comprising: an eigenscale calculation unit that calculates a value of a range blur.   The image is a document image, the document image is compared with a blob area included in the blob image, and a circumscribed rectangular area of a character or a character string included in the blob area is converted into an image element on a specific scale. The image processing apparatus according to claim 1, further comprising an output unit that outputs the area.   The output unit compares the centroid of the character included in the blob area or the centroid of each character of the character string with the blob area, and the circumscribed rectangle of the character or character string including the centroid in the blob area The image processing apparatus according to claim 2, wherein the area is output as an image element area on a specific scale.   The output unit compares a blob area of each blob image generated from a scale image having a different degree of blurring with the document image, thereby obtaining a character area, a word area, a line area, and a character block. The image processing apparatus according to claim 2, wherein the region is output. An image processing method executed by a computer,
Generate multiple scale images with different degrees of blur from the image,
In each scale image, each pixel value in the scale image is compared with a predetermined threshold value, and the scale image is binarized to generate a blob image including a plurality of blobs.
Based on the relationship between the value indicating the degree of blurring of the image and the area or number of each blob, the ratio of the change in the average area or number of blobs to the change in the value indicating the degree of blurring of the image is a predetermined value. An image processing method comprising calculating a degree of blurring in a range that is less than a threshold value. On the computer,
Generate multiple scale images with different degrees of blur from the image,
In each scale image, each pixel value in the scale image is compared with a predetermined threshold value, and the scale image is binarized to generate a blob image including a plurality of blobs.
Based on the relationship between the value indicating the degree of blurring of the image and the area or number of each blob, the ratio of the change in the average area or number of blobs to the change in the value indicating the degree of blurring of the image is a predetermined value. An image processing program for executing a process of calculating a degree of blur in a range that is less than a threshold.