JP2020072325A - Image processing apparatus - Google Patents

Abstract

To solve the problem in which, while, in detecting a boundary line between an original area and a background area, conventionally, there is a method of determining certainty of the boundary line by comparing a distance between a detected boundary line and an original shadow, the conventional method has a problem in that accuracy of the boundary line cannot be correctly determined when the original image is distorted.SOLUTION: An image processing apparatus includes: a document edge detection unit for detecting an edge of a document area from image data; a boundary line calculating unit for calculating a boundary line of the document area based on coordinates of the document edge detected by the document edge detecting unit; a document shadow detection unit for detecting a shadow of the document from the edge of the document; a distance determination unit for determining a distance between a result detected by the document shadow detection unit and a boundary line calculated by the boundary line calculation unit; and a certainty determination unit for determining certainty of the boundary line according to a determination result of the distance determination unit.SELECTED DRAWING: Figure 5

Description

  The present invention relates to an image processing apparatus, and more particularly to detection of a boundary line between a document area and a background area included in image data.

  In recent years, as an image processing apparatus such as a scanner that electronically reads an original, an apparatus that automatically detects the original size and reads the image of the original in an environment in which originals of different sizes are mixed is used. ing. In such an image processing apparatus, in order to detect the document size, in the image data obtained by reading the document, a region where the document is reflected (hereinafter referred to as document region) and a region where the document is not reflected (hereinafter referred to as background region). ) Is being detected. The image processing apparatus performs, for example, the following processing in detecting the boundary. First, the image processing apparatus acquires an image including a document area. Next, the pixels assumed to be the document edge (hereinafter, document edge pixels) are detected from the acquired image, and the coordinate values thereof are acquired. Finally, the boundary is determined by applying a method such as a least square method or Hough transform to the coordinate value of the document edge pixel.

  By the way, in the detection of the document area, erroneous detection may occur. For example, when noise is generated in the image data, the noise may be detected as a document end pixel, and thus a wrong document area may be detected. Further, when the difference in density between the original area and the background area where the original is not reflected is small, the boundary between the original area and the background area is not detected as an original end pixel, and thus an incorrect original area may be detected. ..

  Patent Document 1 proposes a method for detecting the occurrence of such false detection when such false detection occurs. In the technique of Patent Document 1, first, a straight line that is a candidate for a boundary of a document area (hereinafter, a boundary candidate straight line) is calculated from the document edge pixels, and then the calculated boundary candidate straight line and the shadow of the document that appears around the document. Generate a histogram of the distances to and. Then, by analyzing the shape of the generated histogram, it is determined whether the calculated boundary candidate straight line is correct as the boundary between the original area and the background area. For example, when the boundary candidate straight line is a correct boundary, the peak of the histogram becomes sharp because the distance between the boundary candidate straight line and the shadow is almost constant (FIG. 1A). However, when the angle of the boundary candidate straight line is incorrect, the distance between the boundary candidate straight line and the shadow changes continuously, so the peak of the histogram becomes wide (FIG. 1B). Therefore, the correctness of the boundary candidate straight line can be determined from the peak width of the histogram.

2014-72626

  The technique of Patent Document 1 has a problem in that when a document image is distorted, false detection cannot be correctly detected. When a document is distorted and conveyed in a document reading device that reads the document while it is being conveyed, or when a spread book is scanned, the document image may be fan-shaped. In such an image, even if the boundary candidate straight line is appropriate, the distance between the boundary candidate straight line and the shadow continuously changes, so that the peak width of the histogram becomes wide (FIG. 2). Since this histogram has the same shape as the histogram when the angle is wrong, the boundary candidate straight line is determined to be incorrect.

  The present invention is intended to solve the above problem, and an object of the present invention is to realize a determination method capable of correctly determining the boundary between the original area and the background area even if the original image is distorted.

  In order to solve the above problems, the image processing apparatus of the present invention, based on the document edge detection unit that detects the edge of the document area from the image data, and the coordinates of the document edge detected by the document edge detection unit, A boundary line calculation unit that calculates a boundary line of the document area, a document shadow detection unit that detects a document shadow from the edge of the document, a result detected by the document shadow detection unit, and a boundary line calculation unit It is characterized by comprising a distance determining unit for determining a distance to a boundary line, and a certainty determining unit for determining the certainty of the boundary line according to a determination result of the distance determining unit.

  According to the present invention, even if a document image is distorted, it can be determined whether the calculated boundary line between the document region and the background region is correct.

Histogram of the distance between the border and the shadow of the original Histogram of the distance between the boundary line and the shadow of the original when the original image is distorted Configuration diagram of an image reading unit in the present invention Configuration diagram of an image reading apparatus in the present invention Flowchart showing the flow of processing in the first embodiment Flowchart showing the flow of document edge pixel determination processing Flowchart showing the flow of document shadow pixel determination processing Diagram showing how to calculate the least squares method Flowchart showing the flow of processing in the second embodiment

(Example 1)
Hereinafter, an example of an embodiment according to the present invention will be described in detail with reference to the drawings.

  FIG. 3 is a configuration diagram of the image reading unit in the present invention. The image reading unit 301 includes a pickup roller 302 for picking up the document D, a feeding roller 303 for feeding the document bundle D picked up by the pickup roller 302 into the apparatus, and a picked-up document bundle D for each sheet. Separating roller 304, a conveying roller 308 for conveying the fed document to a sheet discharge port, a line image sensor 305 for reading image information of the document separated from the document bundle D, and a position where the line image sensor 305 faces each other. And a registration sensor 306 for detecting a document, and a document detection sensor 309. When the image reading is started, the image reading unit 301 takes in the document bundle D by the pickup roller 302 and the feeding roller 303, and separates the document bundle D one by one by the separation roller 304. While the separated document D is being conveyed in the sub-scanning direction, the image formed on the upper surface thereof is read by the line image sensor 305 along the main scanning direction (direction orthogonal to the document conveying direction). After the image is read, the original is ejected outside the apparatus.

  FIG. 4 is a block diagram of the image reading apparatus according to the present invention. The image reading device 401 includes an image reading unit 402 that reads a document, an image processing circuit 403 that performs image processing on the read image, and a RAM 404 that temporarily stores image data output from the image processing circuit 403. , A ROM 405 storing a program used in the image reading apparatus 401, and controlling the entire image reading apparatus according to the program stored in the ROM 405, analyzing image data in the RAM 405, and calculating a boundary between a document area and a background area. And a transfer unit 408 that transfers the image data in the RAM 404 transmitted from the CPU 406 to the external PC 407 and transfers an interrupt for instructing the image reading from the external PC 407 to the CPU 406.

  Note that in the present embodiment, a configuration is considered in which the image reading device is connected to the image reading device through the transfer unit and is connected to an external PC that gives a reading instruction to the image reading device and receives image data. The configuration may be such that an image can be read without connecting to a PC.

  Further, the processing performed by the CPU 406 in the present embodiment may be performed not by the CPU 406 but by an external PC 407 connected via the transfer unit 408 or another information processing apparatus connected to the external PC 407.

  FIG. 5 shows the flow of processing in this embodiment. Here, a method of detecting the boundary of the upper end of the document will be described.

  In step S101, the user issues an original reading instruction to the external PC 407, and the external PC 407 that has received the original reading instruction sends an image reading instruction to the connected image reading apparatus. The image reading device that has received the image reading instruction conveys the document placed in advance on the image reading device and reads the image.

  In step S102, the read image data is subjected to image processing by the image processing circuit 403 and stored in the RAM 404. In this embodiment, it is assumed that the image data shown in FIG. 2 is stored in the RAM 404.

  In step S103, the CPU 406 analyzes the image data stored in the RAM 404 and detects the document edge pixels. In the present embodiment, the density difference between adjacent pixels is calculated from the edge of the image toward the center of the image, and when a pixel having a density difference between adjacent pixels is equal to or greater than a predetermined value, the pixel is selected. This is performed by the method of determining the pixels at the document edge. However, this detection method is an example, and any method may be used as long as it is a method of detecting a pixel which is assumed to be an edge portion of a document.

  FIG. 6 shows the flow of the document edge pixel determination processing performed in step S103.

  In step S201, the CPU 406 sets 0 in the variables x and y for indicating the coordinates, which are stored in the RAM 404.

In step S202, the threshold value T ₁ for determining whether the density difference is equal to or larger than a predetermined value is acquired. In the present embodiment, the threshold value T ₁ is assumed to be stored in the ROM 405 in advance, and the acquisition is performed by the CPU 406 reading this value.

  In step S203, the CPU 406 determines whether the value of the variable x stored in the RAM 404 is smaller than the number of pixels of the image in the main scanning direction. If it is smaller than the number of pixels in the main scanning direction, the process proceeds to step S204. If the number of pixels is greater than or equal to the number of pixels in the main scanning direction, the process ends.

  In step S204, the CPU 406 determines whether the value obtained by adding 1 to the variable y stored in the RAM 404 is smaller than the number of pixels in the sub-scanning direction of the image. If it is smaller than the number of pixels in the sub-scanning direction, the process proceeds to step S205. If it is equal to or larger than the number of pixels in the sub-scanning direction, the process proceeds to step S208.

In step S205, the CPU 406 calculates the density difference between the pixel at coordinates (x, y) and the pixel at coordinates (x, y + 1), and determines whether the density difference is equal to or greater than the threshold value T ₁ acquired at step S202. .. If the density difference is the threshold value above _{T 1,} the process proceeds to step S207, if thresholds _{T 1} is smaller than, the process proceeds to step S206. The notation "coordinate (x, y)" means that the coordinate in the main scanning direction is x and the coordinate in the sub scanning direction is y. Further, in the present embodiment, the upper left corner of the image is set as the coordinate origin (0, 0), and the coordinate system is determined so that the right side of the image has a larger main scanning coordinate and the lower side of the image has a larger sub scanning coordinate. ..

  In step S206, the CPU adds 1 to the variable y stored in the RAM 404.

  In step S207, coordinate data indicating coordinates (x, y) is added to the document edge pixel storage area in the RAM 404. However, the method of storing the document edge pixels in the RAM 404 may be a general method such as a list structure as long as a plurality of coordinate information can be stored.

  In step S208, the CPU adds 1 to the variable x stored in the RAM 404 and sets 0 to the variable y.

  Through the procedure from step S201 to step S208, the document edge pixels are stored in the document edge pixel storage area in the RAM 404.

  In step S104, the CPU 406 analyzes the image data stored in the RAM 404 and detects the coordinates of pixels corresponding to the shadow of the document. In this embodiment, the density of a pixel is extracted from the edge of the image toward the center of the image, and when a pixel having a density equal to or lower than a certain level is found, the pixel is determined to be the shadow of the document. However, this detection method is an example, and any method may be used as long as it is a method of detecting a pixel assumed to be a shadow of a document.

  FIG. 7 shows the flow of the document shadow pixel determination processing performed in step S104. In step S301, the CPU 406 sets 0 in variables x and y stored in the RAM 404 for indicating coordinates.

In step S302, a threshold value T ₂ for determining whether the density is equal to or higher than a predetermined value is acquired. In this embodiment, the threshold value T ₂ is stored in the ROM 405 in advance, and the acquisition is performed by the CPU 406 reading this value.

  In step S303, the CPU 406 determines whether the value of the variable x stored in the RAM 404 is smaller than the number of pixels of the image in the main scanning direction. If it is smaller than the number of pixels in the main scanning direction, the process proceeds to step S304. If the number of pixels is greater than or equal to the number of pixels in the main scanning direction, the process ends.

  In step S304, the CPU 406 determines whether the value of the variable y stored in the RAM 404 is smaller than the number of pixels of the image in the sub-scanning direction. If it is smaller than the number of pixels in the sub-scanning direction, the process proceeds to step S305. If the number of pixels in the sub-scanning direction is greater than or equal to the number, the process proceeds to step S308.

In step S305, the CPU 406 acquires the density of the pixel at the coordinates (x, y) and determines whether the density is equal to or higher than the threshold T ₂ acquired in step S302. If the density is equal to or higher than the threshold T ₂ , the process proceeds to step S307, and if it is lower than the threshold T ₂ , the process proceeds to step S306.

  In step S306, the CPU adds 1 to the variable y stored in the RAM 404.

  In step S307, coordinate data indicating coordinates (x, y) is added to the document shadow pixel storage area in the RAM 404. However, the method of storing the document shadow pixels in the RAM 404 may be a general method such as a list structure as long as it can store a plurality of coordinate information.

  In step S308, the CPU adds 1 to the variable x stored in the RAM 404 and sets 0 to the variable y.

  By the procedure of steps S301 to S308, the shadow pixels of the document are stored in the document shadow pixel storage area in the RAM 404.

In step S105, the CPU 406 calculates a document boundary candidate (hereinafter, a boundary candidate) based on the coordinate data stored in the document edge pixel storage area. In this embodiment, a quadratic curve is used as a boundary candidate. The quadratic curve may be calculated by any method, but in this embodiment, the least squares method is used. That is, the CPU 406 calculates a quadratic curve that minimizes the squared error from the coordinate data stored in the document edge pixel storage area. In the present embodiment, it is assumed that y = a ₂ x ² + a ₁ x + a ₀ is calculated as the boundary candidate. However, a ₀ , a ₁ , and a ₂ are constants determined by the method of least squares.

Here, the calculation method of the least squares method will be briefly described with reference to FIG. Here, for the sake of simplicity, consider a case where only 10 pieces of coordinate data are stored in the document edge pixel storage area. However, in general, the number of coordinate data stored in the document edge pixel storage area may be larger or smaller than 10. In FIG. 8, these 10 pieces of coordinate data are shown as p ₁ to p ₁₀ . In the least squares method of calculating a quadratic curve, a squared error Σ ¹⁰ _{n = 1} (y _n −a ₂ x _n ² −a _{1 xn} −a ₀ ) ² is calculated for these coordinate data, and this squared value is calculated. Find a ₀ , a ₁ , and a ₂ that minimize the error. In FIG. 8, the calculated quadratic curve is drawn with a broken line.

In step S106, the distance to the boundary candidate calculated in step S105 is calculated for each coordinate data stored in the document shadow pixel storage area, and a distance histogram is generated. The distance from each coordinate data stored in the document shadow pixel storage area to the boundary candidate may be calculated by any method. For example, the following method may be used. First, one coordinate data is acquired from the document shadow pixel storage area. Here, it is assumed that the coordinate data indicates coordinates (x _S , y _S ). Next, the sub-scanning direction coordinate value y _L that the calculated boundary candidate has in the main scanning direction coordinate value x _S is calculated. In this _embodiment, a ^{_{y L = a 2 x S 2}} + a 1 x S + a 0. Then the absolute value of the difference between the _{y S} and _{y L | y} S -y _{L |} is calculated and the distance of this value from the shadow to the boundary candidates.

  Then, it is determined whether the calculated boundary candidate is a correct boundary based on the width of the peak width of the generated histogram. For this determination, the width W of the histogram is acquired in step S107. In this embodiment, the half width (the interval between two points having a value of 50% of the peak height of the histogram) is used as the width W of the histogram. However, using the half-value width as the width W of the histogram is an example, and any value that characterizes the width of the peak may be used. For example, it may be an interval between two points having a peak height value other than 50%, or may be a section in which a certain ratio of elements among all elements forming the histogram is included.

In step S108, the width of the histogram obtained in step S107, the calculated boundary candidate obtains a threshold T _W for determining whether in the correct boundary. In this embodiment, the threshold T _W is stored in the ROM 405 in advance. In the present embodiment, the threshold value stored in the ROM 405 is used as the threshold value T _W , but this is an example, and another value may be used as long as it can be determined whether or not the peak width is wide. May be used.

In step S109, determines whether based on the width W and the threshold T _W peak of the histogram, is calculated boundary candidates are in the correct boundary. In this embodiment, when the width W is larger than the threshold T _W (W> T _W ), it is determined that the calculated boundary candidate is not a correct boundary. On the other hand, when the width W is less than or equal to the threshold T (W ≦ T _W ), it is determined that the calculated boundary candidate is a correct boundary.

  In the present embodiment, a quadratic curve is used when calculating the boundary candidates, but a cubic or higher-order curve or a curve described by another function such as a trigonometric function or an exponential function may be used. Further, the curve described by the coordinate table stored in the ROM 405 may be used instead of the curve described by the function.

(Example 2)
In this embodiment, not only the correctness of the boundary candidate is determined, but also the optimum boundary is detected.

  In this embodiment, the same image reading unit and image reading device as in the first embodiment are used.

  FIG. 9 shows the flow of processing in this embodiment.

  In steps S401 to S404, the document reading instruction is received, the document is read, the document edge pixels are detected, and the document shadow is detected, as in the first embodiment.

In step S405, the maximum value N _MAX of the degree of the curve to be searched when detecting the optimum boundary, which is stored in the ROM 405 in advance, is acquired. However, the maximum value N _MAX may be dynamically calculated instead of using the fixed value stored in the ROM 405. For example, the number of document edge pixels minus 1 acquired in step S403 may be used as N _MAX . As will be described later, in the present embodiment, the curve is calculated by the least square method. Minimum square method, based on the order of curve data number or more of the underlying, since there is a feature that can not be uniquely determined parameter, to the use of the number -1 original edge pixels as N _MAX is the curve Has the advantage that it can be uniquely determined.

  In step S406, 1 is set to the variable N indicating the order of the curve. The variable N is assumed to be stored in a predetermined location of the RAM 404.

In step S407, the optimum curve storage variable for storing the information of the optimum curve, which is provided in a predetermined location of the RAM 404, is initialized. In this embodiment, as the optimum curve information, a set of the curve order N _BEST , N _BEST +1 polynomial coefficients, and histogram width is used. However, this information is an example, and any information may be used as long as the shape of the curve and the degree of suitability can be specified. Further, in the present embodiment, the variables are initialized by setting 0 to N _BEST .

In step S408, an Nth-order curve serving as a boundary candidate is calculated based on the coordinate data stored in the document edge pixel storage area in step S403. The Nth-order curve may be calculated by any method, but in the present embodiment, the least squares method is used. By the least square method, N order curve _{^{y = a N x N + a}} N-1 x N-1 + ··· + a 1 x + a 0 a N + 1 coefficients _{a N, a N-1,} ···, a 1, a ₀ is calculated.

In step S409, the distance to the boundary candidate calculated in step S408 is calculated for each coordinate data stored in the document shadow pixel storage area in step S404, and a distance histogram is generated. The distance from each coordinate data stored in the document shadow pixel storage area to the boundary candidate may be calculated by any method. For example, the following method may be used. First, one coordinate data is acquired from the document shadow pixel storage area. Here, it is assumed that the coordinate data is (x _S , y _S ). Next, the sub-scanning direction coordinate value y _L that the calculated boundary candidate has in the main scanning direction coordinate value x _S is calculated. In this _embodiment, a ^{_{y L = a N x S N}} + a N-1 x S N-1 + ··· + a 1 x S + a 0. Then the absolute value of the difference between the _{y S} and _{y L | y} S -y _{L |} is calculated and the distance of this value from the shadow to the straight line.

  In step S410, the width of the histogram is acquired as in the first embodiment.

In step S411, it is determined whether or not the optimum curve storage variable has been initialized. If it has not been initialized, the process proceeds to step S412, and if it has been initialized, the process proceeds to step S413. In this embodiment, it is determined that the degree N _BEST of the curve in the optimum curve storage variable is initialized if 0 is set, and is not initialized if a value other than 0 is set.

  In step S412, it is determined whether the boundary candidate calculated in step S408 is more suitable than the curve stored in the optimum curve storage variable. Then, if it is determined to be suitable, the process proceeds to step S413, and if it is determined to be unsuitable, the process proceeds to step S414. In this embodiment, the width of the histogram in the optimum curve storage variable is compared with the width of the histogram acquired in step S410, and if the width of the histogram in the optimum curve storage variable is smaller, it is determined to be unsuitable, If the width of the histogram in the curve storage variable is larger, it is determined to be suitable. In the present embodiment, if the width of the histogram acquired in step S410 is equal to the width of the histogram in the optimum curve storage variable, it is determined that the width is not suitable, but this is an example. You may decide that it is suitable. However, the method of determining that it is not suitable has an advantage of being able to detect a boundary described by a simpler function having a lower order.

In step S413, the information of the boundary candidate calculated in step S408 is stored in the optimum curve storage variable. Here, N is stored in the order N _BEST of the curve, a _{N to} a ₀ calculated in step S408 is stored in the coefficient, and the width calculated in step S410 is stored in the width of the histogram.

In step S414, 1 is added to N. In step S415, it is determined whether N is larger than N _MAX . If N is N _MAX or less, the process proceeds to step S408. If N is larger than N _MAX , the process ends.

  With the above procedure, the optimum boundary is stored in the optimum curve storage variable of the RAM 404 at the time when the processing is completed, that is, the optimum boundary is acquired.

  In the present embodiment, only the Nth-order curve is used as the boundary candidate, but this is an example, and a trigonometric function, an exponential function, or a curve described in the coordinate table stored in the ROM 405 may be used. .. When using these curves, for example, information indicating the correspondence between the value of N and the function to be used is stored in the RAM 404 in advance, and the correspondence stored in the RAM 404 when the curve is calculated in step S408. The function to be used may be determined by referring to.

  In the first and second embodiments, the boundary candidate is determined and detected only for the upper end of the document. However, by using the methods of the first and second embodiments, it is possible to easily detect the boundary candidates at the left end, the lower end, and the right end by rotating the method of defining the coordinate system by 90 degrees, 180 degrees, and 270 degrees. ..

  Further, the detection by the method described in the first and second embodiments may be applied only to a specific end of the document, and the other end may be detected by using the conventional technique. For example, a two-page spread document is likely to be distorted at the end including the closed part of the document (for example, upper and lower ends in the case of a left-right opened document), while on the other hand, the end part that does not include the closed part (for example, a left-right opened document). In some cases, there is a feature that distortion is unlikely to occur at the left and right ends. Further, in the case of a document conveyed with distortion, it is likely that the line image sensor that reads the document is distorted at the orthogonal end portion, and the line image sensor that reads the document is less likely to be distorted at the parallel end portion. There are features. In these cases, the boundary of the end portion where distortion is likely to occur may be detected by the method described in the first and second embodiments, and the boundary of the end portion where distortion is unlikely to occur may be detected by the conventional technique. ..

  In the first and second embodiments, when detecting the boundary of one edge of the document, one boundary is detected by using all of the detected document edge and document shadow. However, this is an example, and the detected document edge and document shadow may be divided into a plurality of groups, and one boundary may be detected from each group. As a method of dividing the detected document edge and document shadow into a plurality of groups, for example, for the element having the minimum coordinate in the sub-scanning direction in the document edge and document shadow, the main scanning direction coordinate is set. A method of classifying large elements (elements on the right side of the image) into one group and classifying elements having small coordinates in the main scanning direction (elements on the left side of the image) into another group may be used.

301 image reading unit 302 pickup roller 303 feeding roller 304 separation roller 305 line image sensor 306 registration sensor 307 facing surface 308 conveying roller 401 image reading device 402 image reading unit 403 image processing circuit 404 RAM
405 ROM
406 CPU
407 External PC
408 transfer unit

Claims (4)

A document edge detection unit that detects the edge of the document area from the image data,
A boundary line calculating unit that calculates a boundary line of the document area based on the coordinates of the document edge detected by the document edge detecting unit;
A document shadow detection unit that detects the shadow of the document from the edge of the document,
A distance determination unit that determines the distance between the result detected by the document shadow detection unit and the boundary line calculated by the boundary line calculation unit;
An image processing apparatus, comprising: a certainty determination unit that determines the certainty of the boundary line according to the determination result of the distance determination unit.   The image reading apparatus according to claim 1, wherein the boundary candidate line calculation unit calculates a curve.   The image reading apparatus according to claim 1, wherein the boundary line candidate calculation unit calculates a curve described by a polynomial expression. 4. The boundary candidate line calculation unit calculates a plurality of candidate lines, and the certainty determination unit determines the most probable candidate line from the plurality of candidate lines, any one of claims 1 to 3. The image reading device according to item 1.

Images