JP4946882B2 - Image processing apparatus, image processing system, image processing method, and program - Google Patents

Description

  The present invention relates to an image processing apparatus, an image processing system, an image processing method, and a program.

  As an image processing technique for detecting a graphic element such as a straight line from a noise mixed image, a technique using “Hough transform” is known. For example, Patent Document 1, Patent Document 2, Non-Patent Document 1, and Non-Patent Document 2 shown below are examples of techniques related to straight line detection using the Hough transform.

A specific example of straight line detection according to Patent Document 1 will be described with reference to FIG. FIG. 42 is an explanatory diagram showing the principle when a linear component y is extracted by Hough transform (Hough transform) on one image.
In the figure, an image is represented as an xy plane, and three points P (A), P (B), and P (C) are extracted on a linear component y indicated by a broken line (upper part of the figure).
First, Hough transformation is performed on these points P. That is, assuming a straight line of all angles passing through each point P (one example in the figure), the orthogonal distance ρ and the angle θ from the origin (0, 0) are calculated for each straight line, and the obtained orthogonal The distance ρ and the angle θ are collected for each point, and each point P is represented as a curve on the ρ-θ plane (middle of the figure).

Here, a straight line on the image is expressed as ρ = x cos θ + y sin θ using a distance ρ from the origin and an angle θ with the x axis. For example, there are an infinite number of straight lines that can pass through the point (x _i , y _i ) on the image. As shown in the middle stage, it becomes a single parameter line and is expressed by a curve of ρ = x _i cos θ + y _i sin θ.

  Next, the ρ-θ plane is divided into a large number of cells specified by ρ and θ, and the number of the curves located in each cell is voted to obtain a vote frequency distribution (lower part of the figure). Thereafter, the maximum point (ρ, θ) is determined from the vote frequency distribution.

That is, as shown in the upper part of FIG. 42, when there are points P (A), P (B), and P (C) arranged on one straight line, the corresponding parameter lines are as shown in the middle part of FIG. The point (ρ ₀ , θ ₀ ) where these three parameter lines overlap (in the example of FIG. 42, (ρ ₀ , θ ₀ ) = (3, 7)) is a straight line parameter to be detected. It will be.

Further, when a procedure (inverse Hough transform) opposite to the Hough transform is performed on the local maximum point (ρ, θ), a linear component y indicated by a broken line is specified in the upper part of the figure. Equation representing a straight line to this detection, the _{ρ 0 = xcosθ 0 + y sinθ} 0. In FIG. 42, the coordinates (3, 7) at which the number of votes 3 is obtained are obtained as maximum points. In other words, the obtained local maximum point (ρ, θ) represents the coordinates at which the curves intersect in the middle part of the figure, and these coordinates are common linear components (slopes and intercepts passing through the points P in the upper part of the figure are equal). That is, it represents a broken line component y). Therefore, if this local maximum value (ρ, θ) is inversely Hough transformed, the linear component y that causes each point P can be obtained.

  In Patent Document 2, as in Patent Document 1, an equation is determined by the inverse Hough transform based on the peak value point in the Hough space (parameter space) using the line extraction method by the Hough transform and the inverse Hough transform. The line indicated by the equation is drawn on the image space (paragraph number 0042).

  Further, in Patent Document 2, the outline of the fingerprint image IG as shown in the upper part of FIG. 43 is finally extracted by using a line extraction method based on the Hough transform and the inverse Hough transform. In order to extract a large number of contour lines such as this fingerprint image, feature points PP as shown in the lower part of FIG. 43 are extracted, and in order to extract a single line based on this, Hough transform and inverse processing are performed. The contour line of the fingerprint image is extracted by performing the Hough transform, extracting the line RL, and repeating these processes numerous times. For this reason, the amount of calculation is often enormous for extracting contour lines.

Further, in Non-Patent Document 1 and Non-Patent Document 2, conversion is performed by a method of taking parameters (γ-ω Hough transform) different from normal Hough transform (ρ-θ Hough transform).
JP 2003-40036 A JP-A-9-134427 Toshikazu Wada, Takahiro Fujii, Takashi Matsuyama “γ-ω Hough Transform: Distortion of ρ-θ Parameter Space and Linearization of Voting Trajectory by Variable Sampling—” IEICE Transactions, D-II Vol. J75-D-II, no. 1, pp. 21-30, January 1992 Toshikazu Wada, Masato Seki, Takashi Matsuyama “High-precision γ-ω Hough transform based on geometric characteristics of digital lines” IEICE Transactions, D-II, Vol. J77-D-II, no. 3, pp. 529-539, March 1994

However, Patent Document 1 and Patent Document 2 have a problem that the possibility that a false line due to noise is mistaken as a true line increases or decreases depending on the coordinates on the image.
In other words, the reason why the possibility that a false line due to noise may be mistaken as a true line is increased or decreased depending on the coordinates on the image is that a group of pixels belonging to one line and a group of nearby pixels belonging to another line This is because the same parameter may be considered depending on the position on the image. In this case, the number of pixels detected in the group of pixels belonging to one line increases, and in some cases, the total number of pixels detected in the group of pixels belonging to one line may be It may greatly exceed the total number of detections of the group of pixels belonging to the line.

For example, if a straight line _{ρ = x 0 cosθ + y 0} sinθ passing through the point _{(x 0, y 0),} (x 0, y 0) in the vicinity of the straight line passing through the _{(x 0 + Δx, y 0} + Δy) is ρ + Δρ = (x ₀ + Δx) cos (θ + Δθ) + (y ₀ + Δy) sin (θ + Δθ), but if approximation is performed by expanding Δx, Δy, Δθ and ignoring the second-order or higher terms, Δρ≈ (Δx + y ₀ Δθ ) Cos θ + (Δy−x ₀ Δθ) sin θ. This means that even if Δx, Δy, and Δθ are constant, if (x ₀ , y ₀ ) is different, that is, Δρ is different if the position on the image is different.

Here, when Δx, Δy, Δθ, and Δρ are small, depending on the position of the point (x ₀ , y ₀ ), in addition to the straight line ρ, a straight line ρ + Δρ is also detected, corresponding to the point (x ₀ , y ₀ ). This means that the parameter line corresponding to (x ₀ + Δx, y ₀ + Δy) may be the same parameter line. That is, it means that a plurality of parameter lines may correspond to one point.
On the other hand, a point (ρ, θ) on one parameter space also corresponds to a straight line ρ + Δρ in addition to a straight line ρ on the image, and a plurality of points on the image correspond to one point on the parameter space. In some cases, these lines correspond.

An example of this defect will be described more specifically with reference to FIGS. FIG. 37 and FIG. 38 show an example in which there are two lines α and β, and three points A, B, and C are detected on the lines. The parameter lines on the parameter space corresponding to each point in this case intersect at different coordinates.
On the other hand, in FIG. 39 and FIG. 40, there are two lines α ′ and β ′ to be detected, and three points A ′, B ′, and C ′ are detected. An example in which the positional relationship is the same as point A, point B, and point C in FIG. In this case, the parameter line corresponding to the point A ′, the parameter line corresponding to the point B ″, and the parameter line corresponding to the point C ′ intersect at the same coordinates.
As described above, due to the above-described problem, there is a situation in which the method of intersecting the parameter lines is different only in the position of the detected point.

  That is, even if the mutual positional relationship between point A, point B, and point C and the mutual positional relationship between point A ′, point B ′, and point C ′ do not change, the positions on the coordinates of the points are different. The intersections of the parameter lines are not the same. In this case, since the three parameter lines on the ρ-θ plane in FIG. 40 have only one intersection, two lines α ′ and β ′ correspond to one intersection in the parameter space portion. Thus, when a line corresponding to this intersection is obtained by inverse Hough transform, it is drawn as one line on the xy plane. Therefore, in practice, one false line is drawn where the line α ′ and the line β ′ are to be drawn.

  Thus, depending on the position on the image, the line on the xy plane calculated based on the intersection of the ρ-θ plane does not always indicate an accurate line. That is, there is a problem that the detection accuracy differs depending on the position on the image, in which the line can be accurately detected at one position on the image, but cannot be accurately detected at another position on the image.

  Further, in Non-Patent Document 1 and Non-Patent Document 2, the parameter space distortion is removed by changing the way of taking parameters (γ-ω Hough transform), but normal Hough transform (ρ-θ Hough transform). There was a problem that the amount of calculation would increase significantly.

In addition, another problem is that the above-mentioned phenomenon, where the possibility of mistaking a false line due to noise as a true line, increases or decreases depending on the coordinates on the image, extends the suppression of this phenomenon to general curves other than straight lines. There is no method to apply.
The reason for this is that image processing apparatuses of the related art as described above often have taken into consideration countermeasures for distortion in the parameter space by focusing only on the correspondence to straight lines.

Still another problem is that even if the direction of the same line changes, it may not be distinguished from a false line due to noise and may not be detected.
Counting the number of parameter line overlaps in the parameter space is equivalent to counting the number of pixels belonging to each line on the image in the discretized parameter space, but counting the line by pixel means that the line is Manhattan distance Therefore, if measured in an analog space before discretization, even if the line has the same length, the length changes depending on the inclination.
For example, the lengths of two straight lines shown in FIG. 41 are the same, but if they are discretized, the number of pixels constituting each straight line is different.
In the related art image processing apparatus as described above, this point has not been considered in the Hough transform. However, in an excessively noisy image, there is an increased possibility that the number of constituent pixels that can be detected is less than the number of false line pixels due to noise and cannot be detected.
An example of this problem will be schematically described with reference to the drawings. FIG. 41 shows an example of a straight line having a length of 8 pixels. The two straight lines L1 and L2 have the same length (equivalent to 8 pixels) when measured in units of pixels, but the left side (L1) is longer than the right side (L2) as is apparent. In this case, when measured by the Euclidean distance, the Euclidean distance ed1 on the left side (L1) is about 11.3 pixels.

(Object of invention)
The present invention has been made to solve the problems of the above-described technology, and the object of the present invention is to detect the detection accuracy independent of the position on the image and to require a huge amount of calculation. It is an object to provide an image processing apparatus, an image processing system, an image processing method, and a program capable of detecting a line without performing the above process.

  In order to achieve the above object, the image processing apparatus of the present invention represents a line that can pass through a pixel in an image by an equation including a specific parameter, and uses the parameter space portion having the parameter of the equation as a coordinate axis. An image processing apparatus that performs a process of extracting a line from an image, and draws each candidate line in the image corresponding to the coordinate value based on a coordinate value taken by a combination of the parameters in the parameter space portion Candidate line drawing control means for performing control, candidate line peripheral area determining means for determining, for each of the candidate lines, passing pixels through which the candidate lines and candidate line peripheral areas around the candidate lines pass, and the candidate lines A candidate line for calculating an inclination weight indicating an inclination of a tangent line of the candidate line for each passing pixel and calculating a total value of the inclination weights for each candidate line Based on the total value of each of the candidate lines calculated by the separate index calculation means and the candidate line identification index calculation means, the one in the parameter space portion corresponding to the one candidate line In the parameter value peripheral region around the coordinate value of the parameter combination, search for the presence or absence of the coordinate value of the other parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line, A real line that performs control to determine one candidate line as a real line when there is no total value of the other candidate lines that exceeds the total value of one candidate line within the parameter value peripheral region And a decision control means.

  An image processing system according to the present invention includes a first image processing apparatus that extracts a line from an input image by a first process, and a second process that is different from the first process from the image. A second image processing device to be extracted, a switching control device that controls switching between the first image processing device and the second image processing device, and the first and first controlled by the switching control device. An output device that outputs an image processing result of one or both of the two image processing devices, wherein the second image processing device includes an equation including a specific parameter for a line that can pass through a pixel in the image. Candidate line drawing which performs control for drawing each candidate line in the image corresponding to the coordinate value based on the coordinate value taken by the combination of the parameters in the parameter space portion with the parameter of the equation as the coordinate axis Control means; Candidate line peripheral region determining means for determining, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass, and an index for identifying the candidate line, the candidate line An inclination weight indicating an inclination of the tangent line is calculated for each passing pixel, and is calculated by a candidate line identification index calculation unit that calculates a total value of the inclination weights for each candidate line, and the candidate line identification index calculation unit Based on the total value of each of the candidate lines, the parameter value surrounding area around the coordinate value of one parameter combination in the parameter space portion corresponding to one candidate line, Search for the presence or absence of coordinate values of other parameter combinations corresponding to the other candidate lines that are greater than the total value of the one candidate line, and the one of the candidate lines within the parameter value peripheral region If there is no the total value of the other of said candidate lines exceed the total value is characterized in that it comprises a real line determination control means for controlling to determine the one the candidate lines as a real line, a.

  In the image processing method of the present invention, a line passing through a pixel in an image is represented by an equation including a specific parameter, and a computer uses the parameter space portion having the parameter of the equation as a coordinate axis to draw a line from the image. An image processing method for performing extraction processing, wherein the computer draws each candidate line in the image corresponding to the coordinate value based on a coordinate value taken by a combination of the parameters in the parameter space portion A candidate line drawing control step for performing, and the computer determines a candidate line peripheral region determining step for each of the candidate lines through which the candidate line and a candidate line peripheral region around the candidate line pass, The computer calculates an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line for each passing pixel, A candidate line identification index calculation step of calculating a total value of the inclination weights for the candidate lines, and the computer calculates the total value of each of the candidate lines calculated in the candidate line identification index calculation step. Based on the parameter value peripheral region around the coordinate value of one of the parameter combinations in the parameter space corresponding to one of the candidate lines, the other of the other above the total value of the one candidate line A search is performed for the presence or absence of coordinate values of other parameter combinations corresponding to the candidate line, and the total value of the other candidate lines that exceeds the total value of the one candidate line is not present in the parameter value peripheral region. In some cases, the present invention includes an actual line determination control step for performing control for determining one candidate line as an actual line.

  The program of the present invention represents a process of extracting a line from the image by using a parameter space part having the parameter of the equation as a coordinate axis, and representing a line that can pass through a pixel in the image by an equation including a specific parameter. Candidate line drawing control that is a computer-executable program and performs control for drawing each candidate line in the image corresponding to the coordinate value based on the coordinate value taken by the combination of the parameters in the parameter space portion A function, a candidate line peripheral region determination function for determining the candidate line and a passing pixel through which the candidate line peripheral region around the candidate line passes, and an index for identifying the candidate line A candidate line identification index for calculating an inclination weight indicating an inclination of a tangent line of the candidate line for each passing pixel and calculating a total value of the inclination weights for each of the candidate lines. Based on the total value of each of the candidate lines calculated by the output function and the candidate line identification index calculation function, one parameter of the parameter space unit corresponding to the one candidate line The parameter value surrounding the coordinate value of the combination is searched for the presence or absence of the coordinate value of the other parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line, and the parameter Real line determination control for performing control to determine one candidate line as a real line when there is no total value of the other candidate lines that exceeds the total value of one candidate line in a value peripheral region And having the computer execute a function including the function.

  According to the present invention, a line corresponding to each parameter is drawn on an image so that the parameter space portion and the image have a one-to-one correspondence, so that the detection accuracy of the line is uniform on the image, This is realized by a simple process of drawing a line on the image according to the parameters, and the number of lines to be drawn is limited in number in the vicinity, exceeding the number of others, and the calculation load is reduced. It is possible to provide an excellent image processing apparatus, image processing system, image processing method, and program that are not available in related technologies.

[Basic configuration of image processing apparatus]
First, the basic configuration of the image processing apparatus will be described. The image processing apparatus of the present invention expresses a line that can pass through a pixel in an image by an equation including a specific parameter, and extracts a line from the image by using a parameter space portion having the parameter of the equation as a coordinate axis. The processing is performed.

  The image processing apparatus includes candidate line drawing control means (for example, FIG. 5) that performs control for drawing each candidate line in the image corresponding to the coordinate value based on the coordinate value taken by the combination of the parameters in the parameter space portion. No. 40a shown in No. 1 and No. 142 shown in FIG.

  In addition, the image processing apparatus is a candidate line peripheral area determining unit (for example, reference numeral 47 shown in FIG. 1) that determines, for each of the candidate lines, the candidate line and passing pixels through which the candidate line peripheral area around the candidate line passes. Etc.).

  Further, the image processing apparatus calculates, for each passing pixel, an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line, and calculates a total value of the inclination weight for each candidate line. Candidate line identification index calculating means for calculating (for example, reference numeral 40d shown in FIG. 1) is included.

  In addition, the image processing apparatus is configured to perform one operation in the parameter space unit corresponding to one candidate line based on the total value of each candidate line calculated by the candidate line identification index calculating unit. In the parameter value peripheral area around the coordinate value of the parameter combination, search for the presence or absence of coordinate values of another parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line. When there is no total value of the other candidate lines that exceeds the total value of the one candidate line in the parameter value peripheral region, control is performed to determine one candidate line as a real line. Line determination control means (for example, reference numeral 51 shown in FIG. 1) is included.

  In such an image processing apparatus, a line is drawn on the image with each parameter, and the number of pixels belonging to the line is counted again. Here, lines are drawn on the image based on the extracted parameters, and the number of constituent pixels is counted again. By re-counting the number of pixels for the detected line, only one line for a set of parameters can be obtained, and adjacently detected pixels are not set to the same parameter.

  Further, in this image processing apparatus, it is possible to prevent a line that is very similar to the vicinity of the detected line from being detected repeatedly.

  Furthermore, in this image processing apparatus, calculation in the case of the Hough transform can be avoided by finding a set of parameters constituting the line to be detected by performing the inverse Hough transform from the beginning without performing the Hough transform.

  Hereinafter, an example of a preferred embodiment in which such an “image processing apparatus” of the present invention is applied to an “image processing system” will be specifically described with reference to the drawings.

[First Embodiment]
(Overall configuration of image processing system)
First, the specific configuration of the image processing system according to the present embodiment will be described from the overall configuration, and then the detailed configuration of each unit will be described. FIG. 1 is a block diagram showing an example of a schematic configuration of the entire image processing system according to the first embodiment of the present invention.

  As shown in FIG. 1, the image processing system 1 according to the first embodiment stores an input device 10 including a video camera, a keyboard, and the like, an image processing device 40 operated by program control, and information. Storage device 20 and an output device 30 including a display unit such as a display device and an image forming unit such as a printing device.

  Here, regarding the data connection such as the data reference between the units and means belonging to different devices in FIG. 1, the entire connection is substituted by one arrow between the devices.

  The input device 10 includes an image input unit 11 for inputting an image obtained by a video camera, a digital camera, a scanner, a microscope, a telescope, an X diagnostic device, an ultrasonic diagnostic device, a radar, a sonar, and the like, a keyboard, a mouse, and a touch panel. And a character input unit 12 for inputting character information.

  What is input to the image input unit 11 may be a two-dimensional still image, a three-dimensional spatial data, a moving image, or a variety of speed fields and magnetic fields. It may be a high-dimensional representation of a physical quantity having a spatial and temporal spread, an image feature quantity obtained by various operations such as convolution with a specific function, and its temporal change.

  The storage device 20 includes a threshold storage unit 21 that stores various thresholds input from the character input unit 12, a parameter space storage unit 22 that stores a parameter space used for line detection, a list of detection candidate lines, and the like. And an image storage unit 23 for holding an input image input from the image input unit 11, an image being processed by each unit in the image processing apparatus 40, and an image of a processing result. The storage device 20 can also be configured by one or a plurality of memories, a hard disk, and other various computers.

  The output device 30 includes an image output unit 31 that displays an image, such as a computer display, a television monitor, a projector, and a printer, and a display control that causes the image output unit 31 to display an image stored in the image storage unit 23 according to a user instruction. Part 32.

  The image processing apparatus 40 represents a binarization processing unit 41 that performs binarization processing on each pixel of the image and extracts a candidate region to be detected, and all lines that can pass through one pixel in the candidate region. A parameter calculation unit 42 that calculates a combination of values of each parameter of the equation (for example, a combination of values of ρ and θ), and a parameter space that has each parameter as a coordinate axis (for example, if there are two parameters ρ and θ) A parameter line drawing unit 43 that draws each set of values calculated by the parameter calculation unit 42 on a two-dimensional coordinate system, or a three-dimensional coordinate system if there are three parameters).

  The image processing apparatus 40 also detects the coordinate position of the intersection where each parameter line on the parameter space intersects and the number of times of overlap (number of times of intersection), and the potential of the intersection based on the number of times of overlap at this intersection. And an overlap count measuring unit 45 for selecting candidates.

  Further, the image processing apparatus 40 includes a leading candidate line drawing unit 46 that draws leading candidate lines on the image, a passage determining unit 47 that calculates the coordinates of pixels passing through the leading candidate lines and images near the leading candidate lines. , A weight calculating unit 48 for calculating the inclination weight for the leading candidate line for each pixel of the leading candidate line, a number measuring unit 49 for measuring the inclination weight and the number of pixels of each pixel of the leading candidate line, and the number measuring unit 49 A real line determination unit 51 that determines a real line based on the measurement result in FIG. 2, a real line image generation unit 52 that draws the determined real line, and a module control unit 53 that controls the execution procedure of each unit. It is configured to include.

(Detailed configuration of each part of the image processing apparatus)
The binarization processing unit 41 refers to the image storage unit 23, and based on the pixel value of the input image, each of the two regions, that is, a region that is a candidate for a straight line or a curve that is a detection target and a region other than that are candidate regions The image is divided into two areas as a background area.

In the binarization processing unit 41, the binarization processing of an image that divides an image into two regions can be performed by various methods described below.
As a first method of binarization processing, for example, a ratio of the total number of pixels in a candidate area is set as a threshold value in advance and stored in the threshold storage unit 21, and the image is binarized based on the threshold value. May be applied.
In addition, as a second method of binarization processing, for example, a histogram is created in which the horizontal axis represents pixel values and the vertical axis represents frequency, and the histogram is assumed to be bimodal and binarized using the valley as a threshold value. Laws may apply.
Furthermore, as a third method of the binarization process, a threshold value is determined so that the variance of the pixel value is minimized within each of the candidate region and the background region, and the variance is increased between the candidate region and the background region. A discriminant analysis method for binarization may be applied.
Further, as a fourth method of the binarization process, a fixed threshold method is adopted in which a threshold value of a pixel value is set in advance and stored in the threshold value storage unit 21 and the image is binarized depending on whether the pixel value is larger than the threshold value. You may apply.
Further, as a fifth method of binarization processing, a dynamic threshold method may be applied in which an image is divided into small areas of a certain size and a P tile method, a mode method, or a discriminant analysis method is used for each small area. .

  The parameter calculation unit 42 calculates, for each pixel in the candidate area on the input image, a combination of values of each parameter of an equation representing all straight lines and curves that can pass through the pixel.

For example, when a straight line is expressed as ρ = x cos θ + y sin θ using a distance ρ from the origin and an angle θ with the x axis, if the coordinates of a pixel in a certain candidate region are (x _i , y _i ), the pixel passes through this pixel. All the obtained straight lines are represented by a curve of ρ = x _i cos θ + y _i sin θ in the parameter space. Since 0 ≦ θ ≦ π, ρ is obtained by changing θ within this range.
Also, for example, when detecting a circle with a center coordinate (a, b) radius r, if the coordinates of a certain candidate region are (x _i , y _i ), all circles that can pass through this pixel are a, b, r. Is represented by a circle of (ax _i ) ² + (b−y _i ) ² = r ² . In this case, for example, r is obtained by changing (a, b).

The parameter line drawing unit 43 performs a process of drawing all combinations of the parameter values on the parameter space, with each parameter being a space having coordinate axes as parameter spaces.
Here, a set of each value set of each parameter of the equation expressing all the lines that can pass for one pixel is defined as a parameter line. In this case, the parameter line drawing unit 43 draws the parameter line as a curve or a straight line as an image corresponding to one pixel on the parameter space. Specifically, as shown in FIG. 2, when each pixel PX1, PX2,... Is performed, each parameter line PL1, PL2,.

For example, the parameter line drawing unit 43 reserves a parameter space in the parameter space storage unit 22 for parameter line drawing, and draws all possible parameter value pairs.
The discretization width and the value range of each coordinate axis in the parameter space are given in advance and held in the threshold storage unit 21, for example.
As a method for drawing a parameter line in the parameter space, for example, the value stored in an element in which nothing is drawn is set to zero, and the parameter line to be drawn and the parameter space (for example, the ρ-θ plane shown in the upper part of FIG. 8) The value is incremented by 1 for an element whose coordinates coincide (for example, the address space of the parameter storage unit power shown in the lower part of FIG. The value to be increased need not be 1 as long as it is constant. A series of elements whose values are thus increased is a parameter line. One parameter line is generated for each pixel in the candidate area. On the contrary, you may pay attention to the element which decreased by decreasing the value of the element.

  The intersection detection unit 44 draws a parameter line for all the pixels in the candidate area, and obtains the coordinates of the point where the parameter lines overlap in the parameter space as an intersection.

  For example, the intersection detection unit 44 refers to the parameter space, reads out the value of each element, and when the value is 2 or more, the parameter space storage unit 22 converts all of the values and coordinates on the parameter space as intersections into an intersection list. Save to.

The overlap count measuring unit 45 is a coordinate whose number of parameter lines overlaps is larger than a threshold given from the outside in advance, and the number of overlaps in a predetermined neighborhood (for example, the intersection peripheral area shown in FIG. 3). A combination of parameters indicated by coordinates that do not have a portion exceeding the value is a potential candidate for a parameter combination of a straight line or a curve to be detected.
That is, when the number of intersections at an intersection is equal to or greater than a specific threshold and there is no neighboring intersection with a greater number of intersections in the vicinity of the intersection, the intersection is determined to be an intersection corresponding to a potential candidate line.

For example, the overlap number measurement unit 45 is a case where the value of an element is greater than a threshold value stored in the threshold value storage unit 21 for an intersection in the intersection list, and other element values exist within a predetermined neighborhood. If there are no more points, the potential candidate line bit indicating that it is a potential candidate line is changed from 0 to 1 in the corresponding parameter combination on the intersection list.
The leading candidate line bit is initialized to 0 when the intersection list is generated. Instead, it may be initialized to 1 and changed to 0 if it is determined as a potential candidate line.

Further, instead of 0 and 1 bits, it may be distinguished by any numerical value other than 0 and 1. The value stored in the threshold value storage unit 21 is referred to, for example, as to the extent to which the search is performed for a portion having a value larger than the self.
In general, in the Hough transform, many similar lines may be detected in the immediate vicinity, but a phenomenon in which a plurality of lines are detected in a concentrated manner is prevented by searching and confirming the vicinity.

The leading candidate line drawing unit 46 determines a combination of parameters (for example, leading candidate areas shown in FIG. 4) in the parameter space and within the distance previously given from the leading candidate in the vicinity of the leading candidate and the leading candidate according to the parameters. Draw straight lines and curves on the image as potential candidate lines.
Specifically, as shown in FIG. 4, for a potential candidate area, a potential candidate line PCL1 corresponding to the intersection PO1, a potential candidate line PCL2 corresponding to the intersection PO2, a potential candidate line PCL3 corresponding to the intersection PO3, and so on. Drawing is sequentially performed in the candidate area CA.

For example, the leading candidate line drawing unit 46 draws a straight line or a curve as a leading candidate line on the image according to the parameter combination that is considered to be a leading candidate line such as a leading candidate line bit is 1 on the intersection list. To do.
As a technique for drawing a strong candidate line, for example, an image area having the same size as the input image is secured in the image storage unit 23 as a strong candidate line image, all pixel values are set to zero, and lines and coordinates drawn from a straight line or a curve equation are used. The value is set to 1 for pixels that match. Further, for example, the value may be changed so that each candidate line can be distinguished. Further, for example, the image may be drawn in a binarized image instead of a leading candidate line image.
In this case, however, the pixel value of the leading candidate line is different from the pixel value of the candidate area and the pixel value of the background area.

  The passage determination unit 47 determines, for each potential candidate line, a pixel within the candidate area on the image through which the potential candidate line passes and a neighborhood within a predetermined range from the potential candidate line (for example, the candidate line peripheral area CA shown in FIG. 5). ), The coordinates of the pixel located at (1) are obtained as passing pixels.

For example, the passage determination unit 47 refers to the leading candidate line image, sets a pixel that forms a leading candidate line, such as a pixel having a pixel value of 1, as a target pixel, and normals of the leading candidate line using the coordinates of the target pixel. For the normal direction, the coordinates of a plurality of pixels whose distance from the leading candidate line is within the threshold range held in the threshold storage unit 21 are obtained (for example, the candidate line peripheral area CA shown in FIG. 5).
If any one of the pixels on the binarized image at the same coordinates as those pixels is in the candidate area, it is regarded as a passing pixel and the coordinates of all the passing pixels are stored in the parameter space storage unit 22 as a passing pixel list. To do.
In addition, for example, it may be determined whether the candidate area is included in the vicinity area of N × N pixels centered on the target pixel, not in the normal direction. At this time, N is assumed to be held in a predetermined threshold storage unit 21.

The weight calculation unit 48 obtains the tangent line of the probable candidate line at the target pixel, and obtains the absolute value of the reciprocal of the cosine (cos θ) of the angle formed between the tangent line and the x axis. If this absolute value is less than or equal to the square root of 2, the absolute value of the reciprocal of the cosine of this angle is the slope weight, and otherwise, the absolute value of the reciprocal of sine (sin θ) is the slope weight.
The inclination weight is calculated for each passing pixel. In addition, the inclination weight is calculated for all passing pixels of all possible candidate lines, and is stored in, for example, the parameter space storage unit 22.
Specifically, as shown in FIG. 6, with respect to each passing pixel relating to the leading candidate line PCL1, the inclination weight g1-1, the inclination weight g1-2,. It calculates sequentially like g2-1, inclination weight g2-2, ....

For example, it is assumed that a potential candidate line is represented by a straight line of ρ = x cos θ + y sin θ using a distance ρ from the origin and an angle θ with the x axis. If the angle between the tangent line of the potential candidate line ρ and the x axis is φ (θ + φ = π / 2), 1 / | cos φ | is less than or equal to √2, that is, π / 4 ≦ φ ≦ π / 2, π For / 2 ≦ φ ≦ 3π / 4, the slope weight is 1 / | cos φ |. When 1 / | cos φ | is larger than √2, that is, when 0 ≦ φ ≦ π / 4 and 3π / 4 ≦ φ ≦ π, the inclination weight is 1 / | sin φ |.
Here, θ can also be used as the inclination weight. Since the angle θ is an angle formed between the normal line of the tangent line and the x-axis, when the inclination weight is 0 ≦ θ ≦ π / 4 or 3π / 4 ≦ θ ≦ π, 1 / | cos θ | In this case, 1 / | sinθ |.
Further, in the case of a curved line as well as a straight line, as shown in FIG. 7, the inclination weight can be calculated in the same manner from the inclination of the tangent line for each passing pixel constituting the leading candidate line PCL3.

The number measuring unit 49 sums the inclination weights of all the passing pixels for each of the probable candidate lines to obtain the number, and stores the result (total value of inclination weights, the number of pixels) in the parameter space storage unit 22.
Specifically, as shown in FIG. 6, the first inclination weight total value Sg1 of all the passing pixels with respect to the leading candidate line PCL1, the second inclination weight total value Sg2 of all the passing pixels with respect to the leading candidate line PCL2,. Calculate as follows.

  The actual line determination unit 51 is a case where the number is greater than a threshold given in advance from the outside, and when there is no other location exceeding the number within a predetermined neighborhood in the parameter space, the potential candidate line is It is determined that the line actually exists on the image (real line).

For example, the actual line determination unit 51 is a case where the total inclination weight is larger than a predetermined threshold value held in the threshold value storage unit 21 and the total of the inclination weights is within a predetermined vicinity in the parameter space. If there is no portion exceeding, the real line bit is changed from 0 to 1 on the intersection list as a real line.
The actual line bit is initialized to 0 when the intersection list is generated. Instead, it may be initialized to 1 and changed to 0 if it is determined to be a real line. Further, it may be multi-bit so that it is distinguished by any numerical value other than 0 and 1 instead of 1 bit indicating 0 and 1.
For example, a value stored in the threshold storage unit 21 is referred to for how close the search is performed. The reason for searching for the neighborhood is the same as in the case of the overlap count measuring unit 45. That is, in general, in the Hough transform, many similar lines may be detected in the immediate vicinity, but a phenomenon in which a plurality of lines are detected in a concentrated manner can be prevented by searching and checking in the vicinity. .

  The real line image generation unit 52 generates an image in which only the real line is drawn as a real line image.

For example, the real line image generation unit 52 draws a line for a combination of parameters in which the real line bit is 1 on an output image in which an area is previously secured in the image storage unit 23 and all pixel values are zero.
As a technique for drawing an actual line, for example, a line is expressed by setting the pixel value of a pixel constituting the line to 1. You may make it easy to distinguish a line by changing a pixel value for every line.
In addition, the image may be drawn on a binarized image instead of the output image. In this case, different pixel values from the candidate area and the background area are given so that they can be distinguished.
Furthermore, you may draw on an input image. In this case, for example, processing may be performed so that the detected line is conspicuous even in the input image expressed by shading by inverting the bits of the input image.

The image processing system 1 configured as described above generally operates as follows.
First, an image given from the input device 10 is held in the image storage unit 23 as an input image.
Subsequently, the binarization processing unit 41 binarizes the image and divides the image into an area that seems to have a straight line or a curve and an area that seems to have no straight line or a curve. The former area is a candidate area, and the latter area is a background area. Specifically, the image storage unit 23 is referred to, and based on the pixel value of the input image, a process of dividing the image into two areas, that is, a candidate area to be detected and a background area, is performed. The binarized image is held in the image storage unit 23 as a binarized image.

  Next, the parameter calculation unit 42 obtains the coordinate position in the parameter space of each parameter line corresponding to each pixel in the candidate region, and holds it in the coordinate position storage region of the parameter space storage unit 22. At this time, calculation is performed for all parameter lines of all pixels in the candidate area. The parameter line drawing unit 43 draws each parameter line on the parameter space secured in the parameter line drawing storage area of the parameter space storage unit 22 based on the coordinate position of the parameter space storage unit 22.

  Then, the intersection detection unit 44 refers to the parameter space storage unit 22 to detect the position of the intersection where each parameter line intersects and the number of times of intersection (number of intersections) at the intersection, and stores the intersection list in the parameter space storage unit 22. Save in the intersection list storage area.

  Subsequently, the candidate number lines are selected by selecting the intersections in the overlapping number measurement unit 44. Based on the selection result, the leading candidate line drawing unit 46 draws a leading candidate line on the image.

  Next, the passage determining unit 47 determines the vicinity region of the promising candidate line as a passing pixel, and performs a process of writing the passing pixel list of the passing pixel in the passing pixel list storage region of the parameter storage unit 22.

  Further, the weight calculation unit 48 calculates inclination weights for all passing pixels of all the potential candidate lines based on the tangent lines of the potential candidate lines.

  The number measuring unit 49 calculates the total value of the inclination weights for all the passing pixels of each of the probable candidate lines, and counts the number (total value of the inclination weights). The real line determination unit 51 determines whether there are more points (combinations of parameter values) in the vicinity (candidate part peripheral region) of the real line, and the real line image generation unit 52 detects the detected real line. Draw on top.

  The output image, the input image, or the binarized image in which the line is drawn in this way is held in the image storage unit 23 and is displayed on the image output unit 31 in accordance with an instruction from the display control unit 32 when designated by the user.

Here, the parameter calculation unit 42, the parameter line drawing unit 43, the intersection detection unit 44, the overlap count measuring unit 45, and the leading candidate line drawing unit 46 of the present embodiment constitute the “candidate line drawing control means 40a”. You can also. Further, the passage determination unit 47 of the present embodiment can also be referred to as “candidate line peripheral region determination means”.
Furthermore, the “candidate line identification index calculating means 40d” may be configured by the weight calculating unit 48 and the number measuring unit 49 of the present embodiment. Further, the actual line determination unit 51 of the present embodiment can also be referred to as “real line determination control means”.
Further, the parameter calculation unit 42 and the parameter line drawing unit 43 according to the present embodiment can constitute the “parameter space drawing unit 40b”.
Furthermore, the “selecting means 40 c” can be configured by the intersection detection unit 44 and the overlap number measurement unit 45 of the present embodiment.
Furthermore, the leading candidate line drawing unit 46 of the present embodiment can also be referred to as “candidate line drawing means”.
Further, the binarization processing unit 41 of the present embodiment can also be referred to as “candidate area extraction means”.

The candidate line drawing control means represents a line that can pass through a pixel in the image by an equation including a specific parameter, and is based on a coordinate value taken by a combination of the parameters in the parameter space portion having the parameter of the equation as a coordinate axis. Then, it is possible to perform control for drawing each candidate line in the image corresponding to the coordinate value.
In addition, the candidate line peripheral area determination unit can determine, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral area around the candidate line pass.
Further, the candidate line identification index calculating means calculates an inclination weight that is an index for identifying the candidate line and that indicates an inclination of a tangent line of the candidate line for each passing pixel, and the inclination weight for each candidate line The total value of can be calculated.
Still further, the real line determination control means is configured to provide the parameter space portion corresponding to one candidate line based on the total value of each of the candidate lines calculated by the candidate line identification index calculation means. In the parameter value peripheral area around the coordinate value of the one parameter combination in the above, the coordinate value of the other parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line The presence / absence is searched, and if there is no total value of the other candidate lines exceeding the total value of the one candidate line in the parameter value peripheral region, the one candidate line is determined as a real line You can do control.

  In this case, the parameter space drawing unit can draw a parameter space graphic element corresponding to the pixel for each of the pixels in the parameter space. The selecting means includes the number of intersections at the intersection where each of the parameter space graphic elements intersects, the number of intersections at other intersections around the intersection, and a preset intersection number threshold value. Based on the above, it is possible to select a potential candidate of the coordinate value of the parameter combination in the parameter space portion corresponding to the candidate line in the image. The candidate line drawing means can draw a corresponding candidate line in the image for each coordinate value in a strong candidate area around the coordinate value of the strong candidate selected by the selecting means.

  Further, the candidate area extracting means can extract a candidate area from which the line can be extracted from the image. In this case, the candidate line drawing control means can draw each of the candidate lines in the candidate area. The real line determination control means can determine a real line from among the candidate lines in the candidate area.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

In the image processing method according to the present embodiment, a line that can pass through pixels in an image is represented by an equation including a specific parameter, and a computer uses the parameter space portion with the parameter of the equation as a coordinate axis to allow the computer to It is intended for processing that performs processing for extracting a line from a target.
In this image processing method, the computer performs control for drawing each candidate line in the image corresponding to the coordinate value based on the coordinate value taken by the combination of the parameters in the parameter space portion. Steps (for example, step S100a shown in FIG. 9).
Further, in the image processing method, the computer determines a candidate line peripheral region determining step for determining, for each candidate line, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass (for example, FIG. 9). Step S107 shown in FIG.
In the image processing method, the computer calculates, for each passing pixel, an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line, and the inclination weight for each candidate line. A candidate line identification index calculating step (for example, step S100d shown in FIG. 9).
Further, in the image processing method, the computer uses the parameter space unit corresponding to one candidate line based on the total value of each of the candidate lines calculated in the candidate line identification index calculating step. In the parameter value peripheral area around the coordinate value of the one parameter combination in the above, the coordinate value of the other parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line The presence / absence is searched, and if there is no total value of the other candidate lines exceeding the total value of the one candidate line in the parameter value peripheral region, the one candidate line is determined as a real line An actual line determination control step (for example, step S110 shown in FIG. 9) for performing control is included.

  More specifically, first, based on the pixel value, the computer uses two regions, a candidate region for a straight line or a curve to be detected, and another region as a candidate region and a background region, respectively. Dividing into regions (step S101) <binarization processing step>.

  Next, the computer calculates, for each pixel in the candidate region, a combination of parameters of an equation expressing all straight lines and curves that can pass through the pixel (step S102) <parameter calculation processing step>.

  Subsequently, the computer draws a space about each of the parameters as an axis as a parameter space, and draws all combinations of the parameters as parameter lines on the space in a curved line or a straight line (step S103) <parameter line drawing Processing step>.

  Further, the computer draws parameter lines for all the pixels in the candidate area, and obtains coordinates of points where the parameter lines overlap in the parameter space as intersections (step S104) <intersection detection processing step>.

  In addition, the computer is a coordinate combination in which the number of overlapping parameter lines is a coordinate that is larger than a threshold given from the outside in advance and a parameter indicated by a coordinate that has no other location exceeding the number of overlapping within a predetermined neighborhood, It is set as a promising candidate for a parameter combination of a straight line or a curve to be detected (step S105) <overlap count processing step>.

  Next, the computer draws a straight line or a curve as an effective candidate line on the image in accordance with the parameters for a combination of parameters in the parameter space and in the vicinity of the effective candidate and within the distance previously given from the effective candidate. (Step S106) <Influential candidate line drawing processing step>.

  Further, the computer obtains the coordinates of each of the potential candidate lines by using, as passing pixels, pixels passing through the potential candidate lines in the candidate area on the image and pixels in the vicinity within a predetermined range from the potential candidate lines ( Step S107) <Passing determination processing step>.

  Subsequently, when the absolute value of the reciprocal of the cosine of the angle formed by the tangent and the x axis on the image is equal to or less than the square root of 2, the computer calculates the absolute value of the reciprocal of the cosine of this angle. The slope weight is used, and otherwise, the absolute value of the inverse of the sine is used as the slope weight, and the slope weight is obtained for each passing pixel (step S108) <weight calculation processing step>.

  Further, the computer calculates the total value of the inclination weights of all passing pixels for each of the probable candidate lines, and counts the number (step S109) <number counting processing step>.

  Next, if the number of the computers is greater than the threshold given from the outside in advance, and there is no other location exceeding the number within a predetermined neighborhood in the parameter space, the probable candidate line is actually Is a real line that exists on the image (step S110) <real line determination processing step>.

  Then, the computer generates an image in which only the real line is drawn as a real line image (step S111) <real line image generation processing step>.

Here, the parameter calculation processing step (S102), the parameter line drawing processing step (S103), the intersection detection processing step (S104), the overlap count measurement processing step (S105), and the leading candidate line drawing processing step (S106) of the present embodiment. ) Can also be referred to as a “candidate line drawing control step (S100a)”. Further, the passage determination processing step S107 of the present embodiment can also be referred to as a “candidate line peripheral region determination step”.
Furthermore, it can also be called a “candidate line identification index calculation step” by the weight calculation processing step (S108) and the number measurement processing step (S109) of the present embodiment. In addition, the actual line determination processing step S110 of the present embodiment can also be referred to as “real line determination control step”.
Further, the parameter calculation processing step (S102) and the parameter line drawing processing step (S103) of this embodiment can also be referred to as a “parameter space drawing step (S100b)”.
Furthermore, it can also be called a “selection step (S100c)” by the intersection detection processing step (S104) and the overlap count measurement processing step (S105) of the present embodiment.
Furthermore, the leading candidate line drawing processing step (S106) of the present embodiment can also be referred to as a “candidate line drawing step”.
Further, the binarization processing step (S101) of the present embodiment can also be referred to as a “candidate area extraction step”.

In the parameter space portion drawing step, a parameter space graphic element corresponding to the pixel can be drawn for each of the pixels in the parameter space portion.
In the selection step, the number of intersections at the intersection where each of the parameter space graphic elements intersects, the number of intersections in the other intersections around the intersection, and a preset intersection number threshold value Based on the above, it is possible to select a potential candidate of the coordinate value of the parameter combination in the parameter space portion corresponding to the candidate line in the image.
In the candidate line drawing step, for each coordinate value in a potential candidate area around the coordinate value of the potential candidate selected in the selection step, a corresponding candidate line can be drawn in the image.
In the candidate area extraction step, a candidate area from which the line can be extracted can be extracted from the image. In this case, in the candidate line drawing control step, each candidate line in the candidate area can be drawn. In the real line determination control step, a real line can be determined from each of the candidate lines in the candidate area.

  As described above, according to the present embodiment, once a line having a large number of intersections in the parameter space is detected by the Hough transform, a line is drawn on the image with each parameter, and pixels belonging to the line are drawn. A powerful candidate line drawing unit, a passage determining unit, and a number measuring unit are provided for recounting the number. Here, lines are drawn on the image based on the extracted parameters, and the number of constituent pixels is counted again.

  By re-counting the number of pixels for the detected line, only one line for a set of parameters can be obtained, and adjacently detected pixels are not set to the same parameter. Since only the line is redrawn with the detected parameter, it can be applied regardless of whether the line is a straight line or not.

  For this reason, a line corresponding to each parameter is drawn on the image so that the line corresponds to the parameter space portion and the image on a one-to-one basis, so that the detection accuracy of the line becomes uniform on the image. This is realized by a simple process of drawing a line on the image in accordance with the parameters, and the number of lines to be drawn is limited to be greater than the number of others in the vicinity, and the calculation load can be reduced.

That is, the line detection accuracy can be made uniform on the image without adding a huge calculation load.
The reason is that a straight line corresponding to each parameter is drawn on the image, so that the parameter space and the image have a one-to-one correspondence.
In addition, this is realized simply by drawing a line on the image, and the line to be drawn is also more than the threshold value, and there are more than the number and number of others in the vicinity. By limiting the number, the calculation load can be reduced.

  Furthermore, in the related technology, for the above phenomenon where the possibility that a false line due to noise is mistaken as a true line increases or decreases depending on the coordinates on the image, the suppression of this phenomenon is extended to general curves other than straight lines. It was not applicable. The reason for this is that in related art image processing apparatuses, countermeasures for distortion in the parameter space are often considered by focusing only on dealing with straight lines.

  On the other hand, in the present embodiment, not only a straight line but also a curved line has a uniform detection accuracy on the image. The reason is that only a simple process of drawing according to the parameters is required to make the lines in the parameter space and the image have a one-to-one correspondence.

Furthermore, in the related art, when the direction of the same line is changed, it may not be distinguished from a false line due to noise and may not be detected. Counting the number of parameter line overlaps in the parameter space is equivalent to counting the number of pixels belonging to each line on the image in the discretized parameter space, but counting the line by pixel means that the line is Manhattan. Since the distance is measured, even if the line has the same length as measured in the analog space before discretization, the length changes depending on the inclination.
For example, the lengths of the two straight lines shown in FIG. 41 are the same, but if they are discretized, the number of pixels constituting each straight line is different. In the related art image processing apparatus, this point has not been considered in the Hough transform. However, in an excessively noisy image, there is an increased possibility that the number of constituent pixels that can be detected is less than the number of false line pixels due to noise and cannot be detected.

An example of this problem will be schematically described with reference to FIG. FIG. 41 shows an example of a straight line having a length of 8 pixels. The two straight lines have the same length when measured in pixel units.
However, as you can see, the left side is longer. On the other hand, when measured by Euclidean distance, the left side is about 11.3 pixels (8 × √2).

  In contrast, in this embodiment, the detection accuracy does not change regardless of the direction of the line. This is because the number of pixels counted up according to the slope of the tangent line is adjusted to a value when the length of the line is measured by the Euclidean distance.

  In addition, the image processing apparatus according to the present embodiment is detected by using the number of overlapping points of the parameter in the parameter space or the number of points constituting the line at the time of the inverse Hough transform, or the number of times or the number adding the weight thereto. It has an overlap count metering unit and a real line determination unit that prevent a line similar to the vicinity of the line from being duplicated and detected. In addition, a weight calculation unit that changes the weight according to the slope of the tangent line is provided.

For example, a line on an image (image space) corresponding to an intersection having the largest number of overlaps should have the maximum Euclidean distance. This is because counting the number of parameter line overlaps in the parameter space corresponds to counting the number of pixels belonging to each line on the image in the discretized parameter space. Therefore, the line where the Euclidean distance is maximized is actually checked by measuring the number of pixels and the inclination weight in the image space for the line on the image and the line that seems to belong to the other lines around the line. Only the real line is regarded as a real line, and other Euclidean distance lines are regarded as not real lines and are excluded.
As a result, a line on the image space can be narrowed down to one for a point on one parameter space. For this reason, it is possible to avoid a situation in which a plurality of lines correspond to one point as in the related art.
Although the case where the number of overlaps is the maximum has been described here, the same applies to the case where the number of overlaps is a specific number.

  The description of each part (means) described above can be interpreted as a computer functionalized by a program together with the function of the program, or a plurality of functions permanently functioning by specific hardware. It can also be interpreted that the device comprising the electronic circuit block is described. Therefore, these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof, and is not limited to any one.

[Second Embodiment]
Next, a second embodiment according to the present invention will be described with reference to FIG. In the following, description of the substantially similar configuration of the first embodiment will be omitted, and only different parts will be described. FIG. 10 is a block diagram showing an example of the second embodiment of the image processing apparatus in the image processing system of the present invention.

  In the first embodiment described above, the Hough transform and the inverse Hough transform are performed. However, in the present embodiment, the parameters constituting the line to be detected by performing the inverse Hough transform from the beginning without performing the Hough transform. It is configured to find a set of prices.

(About configuration)
Specifically, as shown in FIG. 10, the second embodiment operates in accordance with program control with an input device 110 including a video camera and a keyboard, as in the first embodiment. The image processing apparatus 140 includes a storage device 120 that stores information, and an output device 130 such as a display device or a printing device.

  In the second embodiment, the input device 110, the storage device 120, and the output device 130 are the same as those in the first embodiment. The configuration of the image processing apparatus 140 is different. Here, the image processing apparatus 140 will be described.

  As shown in FIG. 10, the image processing apparatus 140 according to the second embodiment includes a binarization processing unit 141, an all candidate line drawing unit 142, a passage determination unit 143, a weight calculation unit 144, The measuring unit 145, the actual line determination unit 146, the actual line image generation unit 147, and a module control unit 148 that controls the execution procedure of each unit are configured.

  An image given from the input device 110 is held in the image storage unit 123 as an input image.

  The binarization processing unit 141 refers to the image storage unit 123, and based on the pixel value of the input image, a region that is a candidate for a straight line or a curve to be detected and other regions are set as a candidate region and a background region, respectively. Divide the image into two regions. The binarized image is held in the image storage unit 123 as a binarized image. The binarization processing unit 141 performs the same operation as in the first embodiment.

The all candidate line drawing unit 142 draws a straight line or a curve as a candidate line on the image according to the parameters for all combinations of parameters in the parameter space with the parameter of the equation representing the line on the image as an axis.
Unlike the leading candidate line drawing unit 46 in the first embodiment, drawing straight lines and curves for all parameters instead of specific parameter combinations is the same as the method for drawing other lines. is there.
Specifically, as shown in FIG. 11, the candidate lines CL1,... Are sequentially drawn for each parameter combination.

For example, the all candidate line drawing unit 142 draws a straight line or a curve as a candidate line on the image according to all the parameters.
As a method of drawing a candidate line, for example, an image area having the same size as the input image is secured in the image storage unit 123 as a candidate line image, all pixel values are set to zero, and a line drawn from a straight line or a curve equation coincides with a coordinate. A value of 1 is set for a pixel to be processed. For example, the value may be changed so that each candidate line can be distinguished. Further, for example, the image may be drawn in a binarized image instead of the candidate line image. However, in this case, the pixel value of the candidate line is different from the pixel value of the candidate area and the pixel value of the background area.

  The passage determination unit 143, the weight calculation unit 144, the number measurement unit 145, the real line determination unit 146, and the real line image generation unit 147 are also drawn by the all candidate line drawing unit 142. The processing contents are the same as in the first embodiment.

  In other words, the passage determination unit 143 obtains the coordinates of each candidate line, with the pixel passing through the candidate line in the candidate area on the image and the pixel in the vicinity within a predetermined range from the candidate line as the passing pixel.

For example, the passage determination unit 143 refers to the image, sets a pixel that forms a candidate line as a pixel having a pixel value of 1 as a target pixel, and obtains a normal line of the candidate line using the coordinates of the target pixel. , The coordinates of a plurality of pixels whose distance from the candidate line is within the threshold range held in the threshold storage unit 121 are obtained.
If any one of the pixels on the binarized image at the same coordinates as those pixels is in the candidate area, it is regarded as a passing pixel and the coordinates of all the passing pixels are stored in the parameter space storage unit 122 as a passing pixel list. To do.
In addition, for example, it may be determined whether the candidate area is included in the vicinity area of N × N pixels centered on the target pixel, not in the normal direction. At this time, N is assumed to be stored in the predetermined threshold storage unit 121 in advance.

The weight calculation unit 144 obtains the tangent of the candidate line at the target pixel, and obtains the absolute value of the reciprocal of the cosine (cos θ) of the angle formed between the tangent and the x axis. If this absolute value is less than or equal to the square root of 2, the absolute value of the reciprocal of the cosine of this angle is the slope weight, and otherwise, the absolute value of the reciprocal of sine (sin θ) is the slope weight.
The inclination weight is calculated for every passing pixel of the candidate line. In addition, the gradient weight is calculated for all passing pixels of all candidate lines, and is stored in the parameter space storage unit 122, for example.

For example, using a distance ρ from the origin of the candidate line ρ and the angle θ with the x-axis, ρ = x
If it is expressed by a straight line of cos θ + y sin θ, this is a tangential normal line, so if the slope weight is 0 ≦ θ ≦ π / 4 or 3π / 4 ≦ θ ≦ π, 1 / | cos θ | In this case, 1 / | sinθ |.

  The number measuring unit 145 sums the inclination weights of all the passing pixels for each candidate line to obtain the number, and stores the result (total value of inclination weights, the number of pixels) in the parameter space storage unit 122.

  The actual line determination unit 146 is a case where the number is greater than a threshold given in advance from the outside, and the number of the other is within a predetermined vicinity in the parameter space (for example, the area around the candidate line point supported in FIG. 12). If there is no portion exceeding, the candidate line is determined to be a line that actually exists on the image (real line).

For example, the real line determination unit 146 is a case where the total inclination weight is greater than a predetermined threshold held in the threshold storage unit 121, and the total of the inclination weights is within a predetermined vicinity in the parameter space. If there is no portion exceeding, the actual line bit is changed from 0 to 1 on the candidate line point list as the actual line, for example.
The actual line bit is initialized to 0 when the candidate line point list is generated. Instead, it may be initialized to 1 and changed to 0 if it is determined to be a real line. Further, it may be multi-bit so that it is distinguished by any numerical value other than 0 and 1 instead of 1 bit indicating 0 and 1.
For example, a value stored in the threshold storage unit 121 is referred to for how close the search is performed.

  The real line image generation unit 147 generates an image in which only the real line is drawn as a real line image.

For example, the real line image generation unit 147 draws a line for a combination of parameters in which the real line bit is 1 on an output image in which an area is reserved in the image storage unit 123 in advance and all pixel values are zero.
As a technique for drawing an actual line, for example, a line is expressed by setting the pixel value of a pixel constituting the line to 1. You may make it easy to distinguish a line by changing a pixel value for every line.
In addition, the image may be drawn on a binarized image instead of the output image. In this case, different pixel values from the candidate area and the background area are given so that they can be distinguished.
Furthermore, you may draw on an input image. In this case, for example, processing may be performed so that the detected line is conspicuous even in the input image expressed by shading by inverting the bits of the input image.

Here, all candidate line drawing units 142 of the present embodiment can also be referred to as “candidate line drawing control means”. Further, the passage determination unit 143 according to the present embodiment can also be referred to as “candidate line peripheral region determination means”.
Furthermore, the “candidate line identification index calculation means 140a” can be configured by the weight calculation unit 144 and the number measurement unit 145 of the present embodiment. Further, the actual line determination unit 146 of the present embodiment can also be referred to as “real line determination control means”.

The candidate line drawing control means represents a line that can pass through the pixels in the image by an equation including a specific parameter, and based on the coordinate value taken by the combination of the parameters in the parameter space portion with the parameter of the equation as the coordinate axis, Control is performed to draw each candidate line in the image corresponding to the coordinate value.
The candidate line peripheral area determining means determines, for each of the candidate lines, the candidate line and a passing pixel through which the candidate line peripheral area around the candidate line passes.
Candidate line identification index calculation means calculates an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line for each passing pixel, and the total value of the inclination weights for each of the candidate lines Is calculated.
The real line determination control means is configured to select one of the parameter space portions corresponding to one candidate line based on the total value of each of the candidate lines calculated by the candidate line identification index calculation means. In the parameter value peripheral area around the coordinate value of the parameter combination, search for the presence or absence of coordinate values of another parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line. When there is no total value of the other candidate lines exceeding the total value of the one candidate line in the parameter value peripheral region, control is performed to determine the one candidate line as a real line.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  First, the computer divides the image into two regions based on the pixel values, with the two regions of the detection target line and curve candidates as the candidate region and the background region, respectively. Step S201) <binarization processing step>.

  Next, the computer draws straight lines and curves as candidate lines on the image for all combinations of parameters on the parameter space with the parameters of equations representing straight lines and curves on the image as axes (step S202). <All candidate line drawing processing steps>.

  Subsequently, for each candidate line, the computer obtains coordinates of pixels passing through the candidate line in the candidate area on the image and pixels in the vicinity within a predetermined range from the candidate line as passing pixels (step S203). ) <Passing determination processing step>.

  If the absolute value of the reciprocal of the cosine of the angle formed by the tangent and the x axis on the image is less than the square root of 2 according to the inclination of the tangent of all candidate lines, the computer calculates the absolute value of the reciprocal of the cosine of this angle. The slope weight is used, and the absolute value of the inverse of the sine is used as the slope weight, and the slope weight is obtained for each pixel through which the candidate line passes (step S204) <weight calculation processing step>.

  Further, the computer calculates the total value of the inclination weights of all the passing pixels for each candidate line, and counts the number (step S205) <number counting processing step>.

  Next, if the number of the computers is greater than the threshold given from the outside in advance, and if there is no other location exceeding the number within a predetermined neighborhood in the parameter space, the candidate line is actually an image. A real line which is an upper line is set (step S206) <real line determination processing step>.

  Then, the computer generates an image in which only the real line is drawn as a real line image (step S207) <real line image generation processing step>.

Here, all candidate line drawing processing step (S202) of the present embodiment can also be referred to as “candidate line drawing control step”. Further, the passage determination processing step S203 of the present embodiment can also be referred to as a “candidate line peripheral region determination step”.
Furthermore, it can also be called a “candidate line identification index calculation step” by the weight calculation processing step (S204) and the number measurement processing step (S205) of the present embodiment. Further, the actual line determination processing step S206 of the present embodiment can also be referred to as “real line determination control step”.

In the candidate line drawing control step, the computer represents a line that can pass through the pixels in the image by an equation including a specific parameter, and the coordinate value taken by the combination of the parameters in the parameter space portion having the parameter of the equation as a coordinate axis is used. Based on this, control is performed to draw each candidate line in the image corresponding to the coordinate value.
In the candidate line peripheral region determination step, the computer determines, for each candidate line, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass.
In the candidate line identification index calculation step, the computer calculates an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line for each passing pixel, and the inclination for each of the candidate lines Calculate the total weight.
In the actual line determination control step, the parameter space unit corresponding to one candidate line based on the total value of each of the candidate lines calculated by the computer in the candidate line identification index calculation step. In the parameter value peripheral area around the coordinate value of the one parameter combination in the above, the coordinate value of the other parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line The presence / absence is searched, and if there is no total value of the other candidate lines exceeding the total value of the one candidate line in the parameter value peripheral region, the one candidate line is determined as a real line Take control.

As described above, according to the present embodiment, an increase in the amount of calculation is reduced even when the ratio of the binarized candidate area to the entire image is high while achieving the same operational effects as the first embodiment. it can. The reason is that there is a mechanism for finding a set of values of parameters constituting a line to be detected by performing inverse Hough transform from the beginning without performing Hough transform.
Here, the difference between the first embodiment and the second embodiment is the amount of calculation. The smaller the discretization width, the larger the number of parameter sets, and the second embodiment has a larger calculation amount than the first embodiment. Conversely, the larger the discretization width, the smaller the image space calculation amount.

  Further, when the ratio of the binarized candidate area to the entire image is small, the parameter space is efficiently narrowed down and the amount of calculation is reduced in the first embodiment, but when the ratio is large, the first In the embodiment, since calculation related to the drawing of the parameter line that is not performed in the second embodiment is necessary, the calculation amount in the first embodiment may be larger. In this case, the processing speed of the second embodiment can be improved.

  As described above, in the image processing apparatus according to the present embodiment, in order to avoid a case where the ratio of candidate areas in the binarized image is large and the amount of calculation in the first Hough transform is large, the Hough transform is not performed from the beginning. Inverse Hough transform is used to find a set of parameters that constitute a line to be detected.

  Other configurations, other steps, and operational effects thereof are the same as those in the first embodiment described above. In the above description, the operation content of each step described above and the components of each unit may be programmed and executed by a computer.

[Third Embodiment]
Next, a third embodiment according to the present invention will be described with reference to FIG. In the following, description of the substantially similar configuration of the first embodiment will be omitted, and only different parts will be described. FIG. 14 is a block diagram showing an example of the third embodiment of the image processing system of the present invention.

  In the first embodiment described above, an example in which one image processing device is configured has been described. However, in the third embodiment, different types of first image processing device and second image are provided. An example in which two processing apparatuses are configured is shown. In addition, the third embodiment includes a switching control device that switches between processing performed by any one of the first image processing device and the second image processing device.

(About configuration)
Specifically, as shown in FIG. 14, the image processing system 200 according to the third embodiment includes an input device 210 that inputs an image, and a first image that is input from the input device 210. A first image processing device 240 that extracts a line by processing and a second processing (image processing operation) that extracts a line from the image input by the input device 210 by a second processing (image processing calculation) different from the first processing. The second image processing device 260, the switching control device 270 that controls switching between the first image processing device 240 and the second image processing device 260, and the first that is controlled by the switching control device 270. And an output device 230 that outputs an image processing result of one of the second image processing devices 240 and 270, a storage device 220 that stores the image, the image processing result, and an operation result in the image processing process, Including It is made.

  As shown in FIG. 15, the first image processing device 240 has the same configuration as the image processing device described in the first embodiment, a binarization processing unit 241, a parameter calculation unit 242, Parameter line drawing unit 243, intersection detection unit 244, overlap number measurement unit 245, leading candidate line drawing unit 246, passage determination unit 247, weight calculation unit 248, number measurement unit 249, real line determination unit 251, a real line image generation unit 252, and a module control unit 253 that controls the execution procedure of each unit.

  As shown in FIG. 16, the second image processing device 260 has the same configuration as the image processing device described in the second embodiment, and a binarization processing unit 261 and an all candidate line drawing unit 262. A passage determination unit 263, a weight calculation unit 264, a number measurement unit 265, a real line determination unit 266, a real line image generation unit 267, and a module control unit 268 that controls the execution procedure of these units. It is configured to include.

The switching control device 270 compares the first calculation amount by the first image processing device 240 with the second calculation amount by the second image processing device 260, and the first image processing device 240 performs all processing. Or whether to be processed by the second image processing device 260.
The amount of calculation is determined by the discretization width of the parameter space part and the ratio of binarized image candidate areas.

Here, the switching control device 270 will be described in detail.
As shown in FIG. 17, the switching control device 270 includes a binarization processing unit 271, a parameter calculation calculation amount estimation unit 272, a parameter line drawing calculation amount estimation unit 273, and an intersection detection calculation amount estimation unit 274. Count metric calculation amount estimation unit 275, leading candidate line drawing calculation amount estimation unit 276, passage determination calculation amount estimation unit 277, weight calculation calculation amount estimation unit 278, number metric calculation amount estimation unit 279, and real line determination And a calculation amount estimation unit 281.

  Further, as shown in FIG. 17, the switching control device 270 includes an all-candidate line drawing calculation amount estimation unit 282, an all-pass determination calculation amount estimation unit 283, an all-weight calculation calculation amount estimation unit 284, and an all-pieces calculation calculation. An amount estimation unit 285, an all real line determination calculation amount estimation unit 286, a method selection unit 287, and a module control unit 288 that controls the execution procedure of each unit are configured.

  The binarization processing unit 271 operates in the same manner as in the first embodiment, obtains the ratio of candidate regions on the binarized image, gives a discretization width to each axis of the parameter space, and Estimate the computational complexity of the process.

  17, the parameter calculation calculation amount estimation unit 272, the parameter line drawing calculation amount estimation unit 273, the intersection detection calculation amount estimation unit 274, the overlap count metric calculation amount estimation unit 275, and the probable candidate line drawing calculation amount estimation in the switching control device 270 shown in FIG. The unit 276, the passage determination calculation amount estimation unit 277, the weight calculation calculation amount estimation unit 278, the number metric calculation amount estimation unit 279, and the real line determination calculation amount estimation unit 281 each include the first image processing device 240 illustrated in FIG. Parameter calculation unit 242, parameter line drawing unit 243, intersection detection unit 244, overlap number measurement unit 245, leading candidate line drawing unit 246, passage determination unit 247, weight calculation unit 248, number measurement unit 249, actual line determination unit 251 Estimate the amount of computation during processing.

  If the binarization processing unit 271 knows the ratio of candidate areas on the binarized image and gives the discretization width to each axis of the parameter space, the calculation amount of each process can be estimated.

  The parameter calculation calculation amount estimation unit 272 needs to calculate the number of times for one pixel of the candidate region on the image based on, for example, an equation for drawing a line on the image, a parameter range, and a discretization width of the parameter space. This is approximated and multiplied by the number of pixels in the candidate area to estimate the calculation amount processed by the parameter calculation unit 242 to obtain the parameter calculation calculation amount.

  The parameter line drawing calculation amount estimation unit 273, for example, the number of times that an element is rewritten in the parameter space obtained in the process of the parameter calculation calculation amount estimation unit 272 and the calculation required to rewrite one element. By multiplying the amount, the calculation amount processed by the parameter line drawing unit 243 is estimated and set as the parameter line drawing calculation amount.

  For example, the intersection detection calculation amount estimation unit 274 assumes that the number of pixels in the candidate region is the number of parameter lines, assumes the average length of the parameter lines, and arranges the parameter lines randomly in the parameter space. The number of intersecting points is estimated, the amount of calculation required to obtain the coordinates of one point in the parameter space is multiplied, and the amount of calculation processed by the intersection detection unit 244 is estimated to be the intersection detection calculation amount.

  The overlap count metric calculation amount estimation unit 275, for example, multiplies the number of intersections obtained by the intersection detection calculation amount estimation unit 274 by the calculation amount required to read the value of one intersection and the calculation amount required for comparison of threshold values. The calculation amount processed by the overlap number measurement unit 245 is estimated as the overlap number measurement calculation amount.

  The probable candidate line drawing calculation amount estimation unit 276 exceeds the threshold value, for example, from the number of intersections obtained by the intersection detection calculation amount estimation unit 274 and the average value or distribution of the element values at the intersections obtained simultaneously with the number of intersections. The number of parameters to be a potential candidate line is estimated, the amount of calculation for drawing a single line is estimated from the equation for drawing a line on the image, and this is multiplied by the number of parameters to calculate that the strong candidate line drawing unit 246 processes. The amount is estimated and used as a candidate line drawing calculation amount.

  The passage determination calculation amount estimation unit 277, for example, in the case where the dominant candidate lines are randomly arranged on the binarized image based on the number of parameters to be the estimated leading candidate lines, which is the number of leading candidate lines, is a leading candidate. The number of times that the line passes through the pixels of the candidate area is estimated, and the amount of calculation required to obtain the coordinates of one pixel that passes through is estimated, and the amount of calculation processed by the passage determination unit 247 is estimated and used as the passage determination calculation amount.

  The weight calculation calculation amount estimation unit 278 estimates the calculation amount for calculating the tangent equation from the line equation, for example, and calculates the inclination weight from the calculation amount for calculating the angle between the tangent and the x axis. The calculation amount to be processed is added, and the calculation amount processed by the weight calculation unit 248 is estimated as the weight calculation calculation amount by adding the calculation amount to the average length and the number of leading candidate lines.

  The number measurement calculation amount estimation unit 279 estimates the calculation amount processed by the number measurement unit 249 by multiplying, for example, the average length and the number of leading candidate lines by the calculation amount required to multiply the weight once. Calculated weight.

  For example, the real line determination calculation amount estimation unit 281 compares the total weight of one leading candidate line with a threshold for the estimated number of leading candidate lines and compares the total weight with a predetermined weight. The calculation amount to be processed by the actual line determination unit 251 is estimated by multiplying the required calculation amount to obtain the actual line determination calculation amount.

  17, all candidate line drawing calculation amount estimation unit 282, all passage determination calculation amount estimation unit 283, all weight calculation calculation amount estimation unit 284, all number metric calculation amount estimation unit 285, all real line determination The calculation amount estimation unit 286 performs processing of the all candidate line drawing unit 262, the passage determination unit 263, the weight calculation unit 264, the number measurement unit 265, and the real line determination unit 266 in the second image processing apparatus 260 illustrated in FIG. Estimate time complexity.

The total candidate line drawing calculation amount estimation unit 282 estimates the amount of calculation required to draw a straight line or a curve as a candidate line on the image for all parameter combinations in the parameter space, To do.
For example, the amount of calculation for drawing one line is first estimated from the equation for drawing a line on the image to the number of combinations of all parameters in the parameter space, and all candidate line drawing units 262 process based on this. The amount of calculation is estimated to be the amount of calculation for all candidate line drawing.

The all-passage determination calculation amount estimation unit 283 obtains the coordinates of each candidate line, with the pixel passing through the candidate line in the candidate area on the image and the pixel in the vicinity within a predetermined range from the candidate line as the passing pixel. The amount of calculation required for this is estimated to be the all-pass determination calculation amount.
For example, when the candidate line is arranged on the binarized image, the number of times the candidate line passes through the pixels in the candidate area is estimated, and the amount of calculation required to obtain the coordinates of one pixel passing through is multiplied, and the passage determination unit 263 Estimate the calculation amount to be processed, and set it as the all-pass determination calculation amount.

  The total weight calculation calculation amount estimation unit 284 estimates, for example, a calculation amount for calculating a tangent equation from a line equation, and calculates a calculation amount for determining an angle between the tangent and the x axis, and an inclination weight from the angle. The calculation amount for calculation is added, and it is multiplied by the average length and the number of candidate lines to estimate the calculation amount to be processed by the weight calculation unit 264 as the total weight calculation calculation amount.

The total number calculation calculation amount estimation unit 285 estimates the calculation amount necessary for totaling the slope weights of all the passing pixels for each candidate line, and sets it as the total number calculation calculation amount.
For example, by multiplying the average length and the number of candidate lines by the amount of calculation required to multiply the weight once, the amount of calculation processed by the number counting unit 265 is estimated to be the total number counting calculation amount.

The all real line determination calculation amount estimation unit 286 is a case where the number is greater than a threshold given from the outside in advance, and when there is no other portion exceeding the number within a predetermined neighborhood in the parameter space, the candidate line is The amount of calculation required to be regarded as a real line that is actually present on the image is estimated to be the total real line determination calculation amount.
For example, the total amount of weight of one candidate line is compared with a threshold and multiplied by the amount of calculation required to compare with the total of the weights in a predetermined neighborhood, and the amount of calculation processed by the existing line determination unit 266 is estimated. The actual line judgment calculation amount.

Here, the leading candidate line drawing unit 246 shown in FIG. 15 and the subsequent passage determining unit 247, weight calculating unit 248, number measuring unit 249, and actual line determining unit 251 narrow down the leading candidate lines before these processes. After that, only the strong candidate lines are drawn on the image and the process is continued.
On the other hand, the all candidate line drawing unit 262 shown in FIG. 16 and the subsequent passage determination unit 263, weight calculation unit 264, number measurement unit 265, and actual line determination unit 266 do not narrow down and combine all parameters. Draw a line on the image and continue processing.
As described above, the processing by the first image processing device 240 and the processing by the second image processing device 260 are different depending on whether or not the potential candidate lines are narrowed down.

  First, the method selection unit 287 calculates a parameter calculation calculation amount, a parameter line drawing calculation amount, an intersection detection calculation amount, an overlap count calculation calculation amount, a leading candidate line drawing calculation amount, a passage determination calculation amount, a weight calculation calculation amount, and a number measurement calculation. The sum of the amount and the actual line determination calculation amount is defined as a parameter space calculation amount (first calculation amount).

  The method selection unit 287 also calculates the total of all candidate line drawing calculation amounts, all-pass determination calculation amounts, all weight calculation calculation amounts, all piece metric calculation amounts, and all real line determination calculation amounts as image space calculation amounts (second Calculation amount).

  Further, the method selection unit 287 compares the parameter space calculation amount with the image space calculation amount, and when the parameter space calculation amount is smaller, selects the processing by the first image processing device 240, and When the amount of spatial calculation is smaller, the processing by the second image processing device 260 is selected, and when they are equal, control is performed to select one of the processing based on a predetermined selection constraint.

  When estimating the amount of calculation, the arrangement of lines and the like is determined probabilistically, and the calculation amount itself may have a probability distribution. For example, the estimated value may be added with an error. In this case, when calculating numerical values having a probability distribution, such as when calculating the parameter space calculation amount and the image space calculation amount, the calculated numerical values may have a distribution in consideration of error propagation.

  Here, the parameter calculation calculation amount estimation unit 272, the parameter line drawing calculation amount estimation unit 273, the intersection detection calculation amount estimation unit 274, the overlap count metric calculation amount estimation unit 275, the leading candidate line drawing calculation amount estimation unit according to the present embodiment. 276, the passage determination calculation amount estimation unit 277, the weight calculation calculation amount estimation unit 278, the number measurement calculation amount estimation unit 279, and the real line determination calculation amount estimation unit 281 constitute the “first calculation amount estimation unit 270a”. You can also.

Also, all candidate line drawing calculation amount estimation unit 282, all passage determination calculation amount estimation unit 283, all weight calculation calculation amount estimation unit 284, all number metric calculation amount estimation unit 285, all actual line determination calculation amount of the present embodiment. The estimation unit 286 can constitute the “second calculation amount estimation unit 270b”.
Furthermore, the technique selection unit 287 of the present embodiment can also be referred to as “calculation pattern selection means”.
Furthermore, the binarization processing unit 271, the “first calculation amount estimation unit 270a”, the “second calculation amount estimation unit 270b”, the method selection unit 287, and the module control unit 288 are used to generate “calculation pattern selection control unit”. Can also be configured.

Similarly to the first embodiment, the parameter calculation unit 242, parameter line drawing unit 243, intersection detection unit 244, overlap number measurement unit 245, and leading candidate line drawing unit 246 of the present embodiment can be used as a “candidate”. Line drawing control means "can also be configured. Further, the passage determination unit 247 of the present embodiment can also be referred to as “candidate line peripheral region determination means”.
Furthermore, the “candidate line identification index calculating means” may be configured by the weight calculating unit 248 and the number measuring unit 249 of the present embodiment. In addition, the actual line determination unit 251 of the present embodiment can also be referred to as “real line determination control means”.
Further, the parameter calculation unit 242 and the parameter line drawing unit 243 according to the present embodiment can constitute a “parameter space drawing unit”.
Furthermore, the “selecting means 40 c” can be configured by the intersection detection unit 44 and the overlap number measurement unit 45 of the present embodiment.
Furthermore, the leading candidate line drawing unit 46 of the present embodiment can also be referred to as “candidate line drawing means”.

In this case, the parameter space drawing unit represents a line that can pass through the pixel in the image by an equation including a specific parameter, and the parameter space corresponding to the pixel in the parameter space having the parameter of the equation as a coordinate axis. A graphic element can be drawn for each said pixel.
In addition, the selecting means includes the number of intersections at each intersection where the parameter space graphic elements intersect, the number of intersections at other intersections around the intersection, and a preset number of intersections. Based on the threshold value, it is possible to select a potential candidate for the coordinate value of the parameter combination in the parameter space portion corresponding to the candidate line in the image.
Further, the candidate line drawing means may draw a corresponding candidate line in the image for each coordinate value in a strong candidate area around the coordinate value of the strong candidate selected by the selecting means. it can.

Further, the candidate line peripheral region determining means can determine, for each candidate line, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass.
In addition, the candidate line identification index calculating means calculates an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line for each passing pixel, and the inclination weight for each candidate line The total value of can be calculated.
Further, the actual line determination control means is configured to determine, based on the total value of each of the candidate lines calculated by the candidate line identification index calculation means, the parameter space portion corresponding to one candidate line. Presence / absence of coordinate values of other parameter combinations corresponding to the other candidate lines exceeding the total value of the one candidate line in the parameter value peripheral region around the coordinate values of the one parameter combination Search and control to determine one candidate line as a real line if there is no other total value of the candidate line exceeding the total value of the one candidate line in the parameter value peripheral region It can be carried out.

Further, as in the second embodiment, the “candidate line drawing control means” can be configured by the all candidate line drawing unit 262 of the present embodiment. In addition, the passage determination unit 263 according to the present embodiment can also be referred to as “candidate line peripheral region determination means”.
Furthermore, the “candidate line identification index calculating means” can be configured by the weight calculating unit 264 and the number measuring unit 265 of the present embodiment. Further, the actual line determination unit 266 of the present embodiment can also be referred to as “real line determination control means”.

In this case, the candidate line drawing control means represents a line that can pass through a pixel in the image by an equation including a specific parameter, and a coordinate value taken by a combination of the parameters in the parameter space portion having the parameter of the equation as a coordinate axis. Based on the above, it is possible to perform control for drawing each candidate line in the image corresponding to the coordinate value.
In addition, the candidate line peripheral area determination unit can determine, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral area around the candidate line pass.
Further, the candidate line identification index calculating means calculates an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line for each passing pixel, and the inclination weight of each candidate line is calculated. The total value can be calculated.
In addition, the real line determination control unit is configured to determine, based on the total value of each of the candidate lines calculated by the candidate line identification index calculation unit, in the parameter space unit corresponding to the one candidate line. Presence / absence of coordinate values of other parameter combinations corresponding to the other candidate lines exceeding the total value of the one candidate line in the parameter value peripheral region around the coordinate values of the one parameter combination Search and control to determine one candidate line as a real line if there is no other total value of the candidate line exceeding the total value of the one candidate line in the parameter value peripheral region It can be carried out.

  In addition, the first calculation amount estimation unit is configured to perform drawing processing in the parameter space part by the parameter space part drawing unit of the first image processing apparatus, and selection in the image by the candidate line drawing unit. It is possible to estimate a first calculation amount based on a first calculation pattern when each candidate line is drawn. Furthermore, the second calculation amount estimation means can estimate a second calculation amount based on a second calculation pattern when only drawing processing of each candidate line in the image is performed. Furthermore, the calculation pattern selection unit compares the first calculation amount in the first calculation amount estimation unit with the second calculation amount in the second calculation amount estimation unit, and calculates the calculation amount. The calculation pattern with the smaller amount can be selected.

  The calculation pattern selection control means is configured to perform drawing processing on the parameter space portion by the parameter space portion drawing means and drawing processing of each selected candidate line in the image by the candidate line drawing means. The calculation amount is small based on the first calculation amount based on the first calculation pattern and the second calculation amount based on the second calculation pattern when only drawing processing of each candidate line in the image is performed. You can select the calculation pattern.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  First, based on the pixel value, the computer divides the image into two areas, with the two areas of the detection target line and curve candidates and the other areas as a candidate area and a background area, respectively. A ratio of candidate areas on the binarized image is obtained, and a discretization width is given to each axis of the parameter space to estimate a calculation amount of each processing (step S301) <binarization processing step>.

  Next, the computer estimates the amount of calculation required for calculating a combination of parameters of equations expressing all straight lines and curves that can pass through the pixel in each pixel of the candidate region, (Step S302) <Parameter calculation calculation amount estimation processing step>

  Subsequently, the computer estimates the amount of calculation required to draw a space with each axis as an axis for the parameter as a parameter space and draw all combinations of the parameters as parameter lines on the space in a curved line or a straight line. , Parameter line drawing calculation amount (step S303) <parameter line drawing calculation amount estimation processing step>.

  Then, the computer draws parameter lines for all the pixels in the candidate area, estimates the amount of calculation required to obtain the coordinates of the points where the parameter lines overlap in the parameter space as an intersection, and sets it as the intersection detection calculation amount (Step S304) <Intersection detection calculation amount estimation processing step>.

  Further, the computer detects a combination of parameters indicated by coordinates where the number of overlapping parameter lines is larger than a threshold given from the outside in advance and where there is no other location exceeding the overlapping number within a predetermined neighborhood. Estimate the amount of calculation required to be a promising candidate for the parameter combination of the target straight line or curve, and set it as the overlap count metric calculation amount (step S305) <overlap count metric calculation amount estimation processing step>.

  In addition, for the combination of parameters in the parameter space that are within the distance given from the outside in advance from the potential candidates in the vicinity of the potential candidates in the parameter space, the computer uses straight lines or curves as potential candidate lines on the image according to those parameters. A calculation amount required for drawing is estimated and set as a leading candidate line drawing calculation amount (step S306) <leading candidate line drawing calculation amount estimation processing step>.

  Next, the computer obtains the coordinates of each potential candidate line, using pixels passing through the potential candidate line in the candidate area on the image and pixels in the vicinity within a predetermined range from the potential candidate line as passing pixels. The amount of calculation required to pass is estimated as the passage determination calculation amount (step S307) <passage determination calculation amount estimation processing step>.

  Subsequently, when the absolute value of the reciprocal of the cosine of the angle formed by the tangent and the x axis on the image is equal to or less than the square root of 2, the computer calculates the absolute value of the reciprocal of the cosine of this angle. In other words, the absolute value of the reciprocal of the sine is used as the weight, and the amount of calculation required to obtain the inclination weight for each passing pixel of the leading candidate line is estimated to be the weight calculation calculation amount (step S308) <weight Calculated calculation amount estimation processing step>.

  Then, the computer estimates the amount of calculation required for summing the inclination weights of all the passing pixels for each leading candidate line, and sets it as the number measurement calculation amount (step S309) <number measurement calculation amount Estimation processing step>.

  Further, when the computer is more than the threshold value given from the outside in advance, and there is no other location exceeding the number in the predetermined neighborhood in the parameter space, the probable candidate line is actually A calculation amount required to make a real line which is a line existing on the image is estimated and set as a real line determination calculation amount (step S310) <real line determination calculation amount estimation processing step>.

  Next, the computer estimates the amount of calculation required to draw a straight line or curve as a candidate line on the image for all parameter combinations in the parameter space, and sets it as the total candidate line drawing calculation amount (step S311). ) <All candidate line drawing calculation amount estimation processing step>.

  Subsequently, it is necessary for the computer to obtain the coordinates of each candidate line by using, as a passing pixel, a pixel passing through the candidate line in the candidate area on the image and a pixel in the vicinity within a predetermined range from the candidate line. The calculation amount is estimated and set as an all-pass determination calculation amount (step S312) <all-pass determination calculation amount estimation processing step>.

  If the absolute value of the reciprocal of the cosine of the angle formed by the tangent and the x axis on the image is less than the square root of 2 according to the inclination of the tangent of all candidate lines, the computer calculates the absolute value of the reciprocal of the cosine of this angle. In other words, the absolute value of the inverse of the sine is used as the inclination weight, and the calculation amount required to obtain the inclination weight for each pixel through which the promising candidate line passes is estimated to be the total weight calculation calculation amount (step S313). ) <Total weight calculation calculation amount estimation processing step>.

  Further, for each candidate line, the computer estimates the amount of calculation required for summing the slope weights of all the passing pixels to make the number, and sets it as the total number measurement calculation amount (step S314) <total number measurement calculation amount Estimation processing step>.

  In addition, when the number of the computers is greater than the threshold given from the outside in advance and there is no other location exceeding the number in the predetermined vicinity in the parameter space, the candidate line is actually on the image. The amount of calculation required to be regarded as a real line, which is a line existing in, is estimated as the total real line determination calculation amount (step S315) <all real line determination calculation amount estimation processing step>.

Next, the computer calculates the parameter calculation calculation amount, the parameter line drawing calculation amount, the intersection detection calculation amount, the overlap count metric calculation amount, the leading candidate line drawing calculation amount, the passage determination calculation amount, the weight calculation calculation amount, and the number measurement calculation amount. And the real line determination calculation amount is defined as a parameter space calculation amount (first calculation amount).
In addition, the computer calculates the total of all candidate line drawing calculation amount, all pass determination calculation amount, all weight calculation calculation amount, all piece metric calculation calculation amount, and all real line determination calculation amount (second calculation amount). Amount).
In addition, the computer compares the parameter space calculation amount and the image space calculation amount. If the parameter space calculation amount is smaller than the image space calculation amount, steps S101 to S101 in the first embodiment are performed. A real line image is generated by the processing procedure of S111.
On the other hand, when the image space calculation amount is smaller than the parameter space calculation amount, the computer generates a real line image according to the processing procedure of steps S201 to S207 of the second embodiment.
On the other hand, if the image space calculation amount is equal to the parameter space calculation amount, a real line image is generated by one of the processing procedures based on a predetermined rule (step S316) <method selection processing step>.

  Here, the series of steps in steps S301 to S316 described above can also be referred to as “calculation pattern selection control step”. In the calculation pattern selection control step, the drawing process in the parameter space part by the parameter space part drawing step (step S100b shown in FIG. 9), and the inside of the image by the candidate line drawing step (step S106 shown in FIG. 9). The first calculation amount by the first calculation pattern when performing the drawing process of each selected candidate line, and the second calculation when performing only the drawing process of each candidate line in the image Based on the second calculation amount based on the pattern, it is possible to perform control to select the calculation pattern having the smaller calculation amount.

  As described above, according to the present embodiment, while achieving the same operational effects as the first embodiment, even if the ratio of the binarized candidate area to the entire image is high, the calculation amount is further increased. Can be reduced. The reason for this is that there is a mechanism for finding a set of parameters that constitute a line to be detected by performing inverse Hough transform from the beginning without Hough transform, and by automatically estimating the calculation amount of each method, This is because it can be determined whether or not the Hough transformation should be performed (whether the Hough transformation + inverse Hough transformation or inverse Hough transformation only).

  As described above, in the image processing apparatus according to the present embodiment, the amount of calculation when the Hough transform is performed and then the inverse Hough transform is performed, and the parameters constituting the line to be detected by performing the inverse Hough transform from the beginning without performing the Hough transform. The amount of calculation for finding the set of the two is compared, and it is automatically determined which method has the smaller amount of calculation from the ratio of the candidate areas in the binarized image and the discretization width of the parameter space.

  Other configurations, other steps, and operational effects thereof are the same as those in the first embodiment described above. In the above description, the operation content of each step described above and the components of each unit may be programmed and executed by a computer.

[Fourth Embodiment]
Next, a fourth embodiment according to the present invention will be described with reference to FIG. In the following, description of the substantially similar configuration of the first embodiment will be omitted, and only different parts will be described. FIG. 19 is a block diagram showing an example of the fourth embodiment of the image processing apparatus in the image processing system of the present invention.

  In the present embodiment, an example is disclosed in which a smoothing processing unit is configured before the binarization processing unit of the image processing apparatus of the first embodiment.

(About configuration)
Specifically, as shown in FIG. 19, the image processing apparatus 340 in the image processing system of the fourth embodiment includes a smoothing processing unit 341 that smoothes an image, and the first embodiment. The binarization processing unit 342, the parameter calculation unit 343, the parameter line drawing unit 344, the intersection detection unit 345, the overlap count measurement unit 346, the leading candidate line drawing unit 347, and the passage determination having the same configuration. Unit 348, weight calculation unit 349, number measurement unit 351, real line determination unit 352, real line image generation unit 353, and module control unit 354 that controls the execution procedure of these units. .

  As a smoothing method in the smoothing processing unit 341, for example, a method of convolving a mask whose weight is a two-dimensional Gaussian into an image, a method of taking an average in a local region, a method of taking an intermediate value in a local region, There are various methods such as a method of least square fitting of a multidimensional aspect, and any method may be applied.

  Here, the smoothing processing unit 341 of the present embodiment can also be referred to as “smoothing processing means”. In addition, the parameter calculation unit 343, the parameter line drawing unit 344, the intersection detection unit 345, the overlap count measurement unit 346, and the leading candidate line drawing unit 347 according to the present embodiment provide a “candidate line drawing control unit”. It can also be configured.

  The smoothing processing means can smooth the image before extracting the candidate area. In this case, the candidate line drawing control means can draw the candidate line based on the image of the candidate area smoothed by the smoothing processing means.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In the image processing method according to the present embodiment, the computer performs smoothing processing before binarization processing (step S401) <smoothing processing step>.

  Subsequently, the processing of Step S402 to Step S412 which is the same processing procedure as that of the first embodiment is performed.

  Here, in the smoothing process step of step S401 of the present embodiment, the image is smoothed before extracting candidate areas. In this case, in the candidate line drawing control step (steps S403 to S407), the candidate line can be drawn based on the image of the candidate area smoothed in the smoothing processing step.

  As described above, according to the present embodiment, it is possible to reduce noise by smoothing while exhibiting the same operational effects as those of the first embodiment.

  Other configurations, other steps, and operational effects thereof are the same as those in the first embodiment described above. In the above description, the operation content of each step described above and the components of each unit may be programmed and executed by a computer.

  In the present embodiment, for the sake of simplicity, the configuration example in which the smoothing processing unit is added to the image processing apparatus according to the first embodiment has been described. However, the image processing apparatus according to the second embodiment performs smoothing. When the processing unit is added, a smoothing processing unit may be added to each of the first and second image processing apparatuses of the third embodiment.

[Fifth Embodiment]
Next, a fifth embodiment according to the present invention will be described with reference to FIG. In the following, description of the substantially similar configuration of the first embodiment will be omitted, and only different parts will be described. FIG. 21 is a block diagram showing an example of the fifth embodiment of the image processing apparatus in the image processing system of the present invention.

  The present embodiment discloses an example in which an expansion / contraction processing unit is configured after the binarization processing unit of the image processing apparatus according to the first embodiment.

(About configuration)
Specifically, the image processing apparatus 440 in the image processing system according to the fifth embodiment performs an expansion / contraction process on the binarized image candidate area as many times as given from the outside, as shown in FIG. An expansion / contraction processing unit 442, a binarization processing unit 441, a parameter calculation unit 443, a parameter line drawing unit 444, and an intersection detection unit 445 having the same configuration as that of the first embodiment. , Overlapping number measurement unit 446, leading candidate line drawing unit 447, passage determination unit 448, weight calculation unit 449, number measurement unit 451, real line determination unit 452, real line image generation unit 453, and the like And a module control unit 454 that controls the execution procedure of each unit.

As a technique of the expansion / contraction process in the expansion / contraction processing unit 442, a technique generally used in image processing may be used.
For example, in the dilation process, if there is a pixel in the candidate area in any one of the eight neighborhoods of the target pixel, the target pixel is set as the candidate area.
Further, for example, in the contraction process, if there is a pixel that belongs to the background region in any of the eight neighborhoods of the pixel of interest, the attention is determined to be the background region.

  Here, the expansion process and the contraction process in the expansion / contraction process unit 442 may be performed in a reversed order. Further, the expansion / contraction processing unit 442 may perform a process in which the expansion process and the contraction process are repeated an arbitrary number of times in an arbitrary order.

  Here, the expansion / contraction processing unit 442 of the present embodiment can also be referred to as “expansion / contraction means”. In addition, the parameter calculation unit 443, the parameter line drawing unit 444, the intersection detection unit 445, the overlap number measurement unit 446, and the leading candidate line drawing unit 447 according to the present embodiment provide a “candidate line drawing control unit”. It can also be configured.

  The expansion / contraction processing means can perform one or both of the processes of expanding and contracting the candidate region a specific number of times. In this case, the candidate line drawing control means can draw the candidate line based on the image of the candidate area processed by the expansion / contraction processing means.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In the image processing method according to the present embodiment, first, based on the pixel value, the computer converts a candidate area and a background area into two areas, that is, a candidate area for a straight line or a curve to be detected and another area. Then, the image is divided into two regions (step S501) <binarization processing step>.

  Next, after the binarization manual processing, the computer performs one or both of expansion processing and contraction processing on the candidate region of the binarized image for the number of times given in advance from outside (step S502) < Expansion and contraction processing step>.

  Subsequently, the processing in steps S503 to S512, which is the same processing procedure as in the first embodiment, is performed.

  Here, in the expansion / contraction process step of step S502 of the present embodiment, one or both of the processes for expanding and contracting the candidate area are performed a specific number of times. In this case, in the candidate line drawing control step (steps S503 to S507), the candidate line can be drawn based on the image of the candidate area processed in the expansion / contraction processing step.

  As described above, according to the present embodiment, while exhibiting the same operational effects as the first embodiment, noise is reduced by such expansion processing and contraction processing, and disconnected lines are connected to some extent. You can also

  Other configurations, other steps, and operational effects thereof are the same as those in the first embodiment described above. In the above description, the operation content of each step described above and the components of each unit may be programmed and executed by a computer.

  In the present embodiment, for the sake of simplicity, the configuration example in which the expansion / contraction processing unit is added to the image processing apparatus of the first embodiment has been described. However, the image processing apparatus of the second embodiment has the expansion / contraction When a processing unit is added, when an expansion / contraction processing unit is added to each of the first and second image processing devices of the third embodiment, expansion / contraction is added to the image processing device of the fourth embodiment. It may be a case where a processing unit is added.

[Sixth Embodiment]
Next, a sixth embodiment according to the present invention will be described with reference to FIG. In the following, description of the substantially similar configuration of the first embodiment will be omitted, and only different parts will be described. FIG. 23 is a block diagram showing an example of the sixth embodiment of the image processing apparatus in the image processing system of the present invention.

  In the present embodiment, the image reduction processing unit configured in the preceding stage of the binarization processing unit of the image processing apparatus of the first embodiment, and the parameter for performing the parameter fine adjustment control process by controlling each unit An example in which a fine adjustment control unit is configured is disclosed.

(About configuration)
Specifically, as shown in FIG. 23, the image processing apparatus 540 in the image processing system of the sixth embodiment includes an image reduction processing unit 541, a parameter fine adjustment control unit 554, and the first embodiment. The binarization processing unit 542, the parameter calculation unit 543, the parameter line drawing unit 544, the intersection detection unit 545, the overlap number measurement unit 546, and the promising candidate line drawing unit 547 are configured in the same manner as The passage determining unit 548, the weight calculating unit 549, the number measuring unit 551, the real line determining unit 552, and the real line image generating unit 553 are included.

  The image reduction processing unit 541 reduces the input image at a predetermined magnification, sets the image as a reduced image (first reduced image), obtains a real line, and once the real line is obtained, the first reduced image is obtained. An image reduced to a higher reduction ratio is newly generated to obtain a reduced image (second reduced image).

  The parameter fine adjustment control unit 554 has a module control function for controlling the execution procedure of each unit, regards the actual line once obtained as a new potential candidate line, newly calculates a real line near the parameter space, Control to repeatedly execute image reduction and derivation of a new real line for a predetermined number of times or until the accuracy in the preset parameter space is reached (until the fluctuation range in the parameter space becomes smaller than a predetermined value). I do.

Here, the detailed operation in the image processing apparatus 540 when the image is repeatedly reduced and the reduction rate is changed will be described.
First, following the binarization processing unit 542, the parameter calculation unit 543, the parameter line drawing unit 544, the intersection detection unit 545, and the overlap count measurement unit 546, the processing in the leading candidate line drawing unit 547 and thereafter is performed. The

Next, in the second and subsequent iterations, control is performed so that the process in the leading candidate line drawing unit 547 is performed immediately after the binarization processing unit 542.
This is because, in the first stage (first process), a process for narrowing down the area from the entire parameter space is performed by the Hough transform. Therefore, the parameter calculation unit 543, the parameter line drawing unit 544, the intersection detection unit 545, and the overlap count measurement unit Processing at 546 is required.
On the other hand, in the second and subsequent processing stages, a real line is once obtained at a certain image reduction ratio in the first stage (first process), and the real line is directly searched as a promising candidate line in the vicinity thereof to set parameters. Since it will be refined, processing in the parameter calculation unit 543, the parameter line drawing unit 544, the intersection detection unit 545, and the overlap count measurement unit 546 becomes unnecessary.

  Here, the image reduction processing unit 541 of the present embodiment can also be referred to as “image reduction processing means”. Further, the parameter fine adjustment control unit 554 of the present embodiment can also be referred to as “first adjustment control means”.

  In this case, the image reduction processing means can perform a process of reducing the image. The first adjustment control means determines the real line in the first reduced image obtained by reducing the image at the first magnification by the image reduction processing means, and then determines the determined real line. As a new new candidate line, a new search is performed for the area surrounding the new candidate line around the new candidate line, and the parameter is set in the second reduced image reduced at the second magnification higher than the first magnification. A new real line is newly determined using the space, and the reduction and the determination of the new real line are repeatedly performed in stages, thereby performing adjustment control of the determination of the real line.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In the image processing method according to the present embodiment, first, the computer reduces the input image at a predetermined magnification to make the image a reduced image (step S601) <image reduction processing step>.

  Next, the computer divides the image into two regions based on the pixel values, with the two regions of the detection target line and curve candidates and the other regions as a candidate region and a background region, respectively. (Step S602) <binarization processing step>.

Subsequently, the computer determines whether the process is the first process or the second process or later (step S603) <process count determination process step>.
In this determination process, when the computer determines that the process is the first process, the computer performs the process from step S604, and when the computer determines that the process is the second or later process, the computer Performs the process from step S608.

  And a computer performs the process of step S604-step S613 which is the process sequence similar to the said 1st Embodiment.

Further, the computer determines whether or not the setting accuracy on the preset parameter space has been reached (step S614) <accuracy determination processing step>.
In this determination process, when the computer determines that the setting accuracy in the preset parameter space has not been reached, the computer returns to step S601 and performs the processes of steps S601 to S603 and steps S608 to S613. (Second image reduction processing step, second binarization processing step, second and subsequent processing frequency determination processing step, second and subsequent dominant candidate line drawing processing step, second and subsequent passage determination processing step, The second and subsequent weight calculation processing steps, the second and subsequent number counting processing steps, the second and subsequent real line determination processing steps, and the second and subsequent real line image generation processing steps) are repeatedly executed.
On the other hand, when the computer determines that the setting accuracy in the preset parameter space has been reached, the computer ends the process.

  Here, in the second and subsequent processing stages, a real line is once obtained at a certain image reduction ratio in the first stage (first process), and the real line is directly searched as a promising candidate line in the vicinity thereof, and parameters are set. Therefore, the processing from step S604 to step S607 becomes unnecessary.

  As described above, the computer obtains the real line, and once the real line is obtained, an image reduced to a size larger than the reduced image is newly generated to obtain a reduced image.

  Then, the computer regards the actual line once obtained as a new candidate line or a candidate line, re-determines a new real line in the vicinity of the parameter space, reduces the image with different magnification, and derives a new real line. It is repeatedly executed until the predetermined number of times or the fluctuation range in the parameter space becomes smaller than a predetermined value.

  Here, the step of performing the repetitive control in steps S603 to S614 can also be referred to as a “first adjustment control step”. In the “first adjustment control step”, after the real line is determined by the first reduced image obtained by reducing the image at the first magnification in the image reduction processing step, the real line thus determined is determined. As a new new candidate line, a new search is performed for the area surrounding the new candidate line around the new candidate line, and the parameter is set in the second reduced image reduced at the second magnification higher than the first magnification. A new real line is newly determined using the space, and the reduction and the determination of the new real line are repeatedly performed in stages, thereby performing adjustment control of the determination of the real line.

  As described above, according to the present embodiment, the same effect as the first embodiment can be obtained, but the image is sequentially made high definition, and the line is detected with higher accuracy and lower calculation amount. be able to.

  As described above, the image processing apparatus according to the present embodiment includes the image reduction unit that reduces the image and the parameter fine adjustment unit that reduces the image in stages and gradually improves the position accuracy of the detected line.

  Other configurations, other steps, and operational effects thereof are the same as those in the first embodiment described above. In the above description, the operation content of each step described above and the components of each unit may be programmed and executed by a computer.

  In the present embodiment, for the sake of simplicity, the configuration example in which the image reduction processing unit and the parameter fine adjustment control unit are added to the image processing apparatus of the first embodiment has been described. When an image reduction processing unit and a parameter fine adjustment control unit are added to the image processing device, an image reduction processing unit and a parameter fine adjustment control unit are provided in each of the first and second image processing devices of the third embodiment. In the case of adding the image reduction processing unit and the parameter fine adjustment control unit to the image processing device of the fourth embodiment, the image reduction processing unit and the parameter fine adjustment are added to the image processing device of the fifth embodiment. It may be a case where an adjustment control unit is added.

(Modification of the sixth embodiment)
(About configuration)
FIG. 25 is a block diagram illustrating a configuration example in which an image reduction processing unit and a parameter fine adjustment control unit are added to the second embodiment.

  As shown in FIG. 25, the image processing apparatus 640 in the image processing system is different from the second embodiment in place of the all candidate line drawing unit in addition to the image reduction processing unit 641 and the parameter fine adjustment control unit 649. All candidate line leading candidate line drawing units 643 are provided.

  The all candidate line leading candidate line drawing unit 643 draws lines on the image for all combinations of parameters on the parameter space after first forming the first reduced image (the first time all candidate line drawing processing). However, once a real line is obtained and a real line is obtained with a reduced image (second reduced image) of the next size, a line is drawn only for a combination of parameters in the vicinity of the found real line in the parameter space. (From the second time on, a powerful candidate line drawing process).

  In addition, the image processing device 640 has the same configuration as that of the second embodiment, a binarization processing unit 641, a passage determination unit 644, a weight calculation unit 645, a number measuring unit 646, a real line The determination unit 647 and the real line image generation unit 648 are included.

  In the image processing apparatus 640 having the above-described configuration, the parameter fine adjustment control unit 649 reduces the input image at a predetermined magnification by the image reduction processing unit 641 following the binarization processing unit 641 and reduces the input image. Control is performed to change the image size to be a reduced image (first reduced image).

  Thereafter, the parameter fine adjustment control unit 649 performs processing in all candidate line leading candidate line drawing units 643 and subsequent processing (processing in the first pass determination unit 644, processing in the first weight calculation unit 645, and first time Control at the number measuring unit 646 of the first actual line determination unit 647).

  Here, in the first stage (first processing), all candidate line leading candidate line drawing units 643 draw all lines on the image for all parameter combinations in the parameter space. I do.

  Next, the parameter fine adjustment control unit 649 obtains the actual line, and once the actual line is obtained, the image reduction processing unit 641 similarly uses the second reduction process and after that it is higher than the first reduced image. An image reduced to the reduction magnification is newly generated to obtain a reduced image (second reduced image), and processing in all candidate line leading candidate line drawing units 643 and subsequent processing (processing in the second and subsequent pass determination units 644) Control is performed so that processing in the second and subsequent weight calculation unit 645, processing in the second and subsequent number measurement unit 646, and processing in the second and subsequent actual line determination unit 647) is performed.

  Here, in the second and subsequent processing stages, the actual line is obtained once using the parameter space at the image size (first reduced image) which is the first stage (first processing), and the next image is obtained. When obtaining a real line with the size (second reduced image), in the parameter space, a line is drawn for only the combination of parameters when the found real line is directly searched as a potential candidate line in the vicinity, and all candidate lines are drawn. The leading candidate line drawing unit 643 performs leading candidate line drawing processing by refining the parameters.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In the image processing method according to the present embodiment, first, the computer reduces the input image at a predetermined magnification and sets the image as a reduced image (first reduced image) (step S601) <Image reduction processing step> .

  Next, the computer divides the image into two regions based on the pixel values, with the two regions of the detection target line and curve candidates and the other regions as a candidate region and a background region, respectively. (Step S702) <binarization processing step>.

Subsequently, the computer performs an all candidate line leading candidate line drawing process step of Step 703.
Here, it is determined whether the process is the first process or the second and subsequent processes <process count determination process step>. In this determination process, if the computer determines that this is the first process, the computer draws lines on the image for all parameter combinations in the parameter space. In the line drawing process, all candidate line drawing processes are performed.

  On the other hand, when the computer determines that the process is the second and subsequent processes, the computer temporarily uses the parameter space at the image size (first reduced image) that is the first stage (first process). In order to obtain a real line at the next image size (second reduced image), the real line found in the parameter space is directly used as a strong candidate line in the parameter space in the candidate line drawing process. A line is drawn only for a combination of parameters when searching in the vicinity, and the parameters are refined to perform a powerful candidate line drawing process.

  Then, the computer performs the processing from step S704 to step S708, which is the same processing procedure as in the second embodiment.

Further, the computer determines whether or not the setting accuracy in the preset parameter space has been reached (step S709) <accuracy determination processing step>.
In this determination process, when the computer determines that the setting accuracy in the preset parameter space has not been reached, the computer returns to step S701 and performs the processes of steps S701 to S708 (second and subsequent images). Reduction processing step, second and subsequent binarization processing steps, second and subsequent all-candidate candidate line drawing processing steps, second and subsequent passage determination processing steps, second and subsequent weight calculation processing steps, second and subsequent processing steps The number measurement processing step, the second and subsequent real line determination processing step, and the second and subsequent real line image generation processing step) are repeatedly executed.
On the other hand, when the computer determines that the setting accuracy in the preset parameter space has been reached, the computer ends the process.

  Here, in the second and subsequent processing stages, a real line is once obtained using the parameter space at the image size (first reduced image) which is the first stage (first process), and the real line is The parameter is refined by searching in the vicinity as a strong candidate line as it is.

  As described above, the computer obtains the real line, and once the real line is obtained, an image reduced to a size larger than the reduced image is newly generated to obtain a reduced image.

  Then, the computer regards the actual line once obtained as a new candidate line or a candidate line, re-determines a new real line in the vicinity of the parameter space, reduces the image with different magnification, and derives a new real line. It is repeatedly executed until the predetermined number of times or the fluctuation range in the parameter space becomes smaller than a predetermined value.

[Seventh Embodiment]
Next, a seventh embodiment according to the present invention will be described with reference to FIG. In the following, description of the substantially similar configuration of the first embodiment will be omitted, and only different parts will be described. FIG. 27 is a block diagram showing an example of the seventh embodiment of the image processing apparatus in the image processing system of the present invention.

  In the present embodiment, a parameter space coarse-graining processing unit configured after the binarization processing unit of the image processing apparatus of the first embodiment, and a parameter fine-tuning control process by controlling each unit The example which comprised the parameter fine adjustment control part which performs is disclosed.

(About configuration)
Specifically, as shown in FIG. 27, the image processing apparatus 740 in the image processing system of the seventh embodiment includes a parameter space coarse-grain processing unit 742, a parameter fine adjustment control unit 754, and the first The binarization processing unit 741, the parameter calculation unit 743, the parameter line drawing unit 744, the intersection detection unit 745, the overlap count measurement unit 746, and the potential candidate line drawing, which have the same configuration as that of the first embodiment. A unit 747, a passage determination unit 748, a weight calculation unit 749, a number measuring unit 751, a real line determination unit 752, and a real line image generation unit 753.

  The parameter space coarse graining processing unit 742 obtains a real line by increasing the discretization width of the parameter space to a predetermined width to obtain a real line, and once the real line is obtained, the parameter space is more densely discrete. Perform processing.

  The parameter fine adjustment control unit 754 has a module control function for controlling the execution procedure of each unit, re-determines a real line in the vicinity of the real line once obtained in the subdivided parameter space, and has different discretization widths. Control is performed to repeatedly execute the derivation of a new real line in the space a predetermined number of times or until the discretization width reaches a predetermined value.

  Here, a detailed operation in the case where the discretization width of the parameter space is repeatedly reduced stepwise in the image processing device 740 will be described.

  First, the parameter fine adjustment control unit 754 controls the parameter space coarse-grain processing unit 742 to change the discretization width of the parameter space following the binarization processing unit 741.

  Thereafter, the parameter fine adjustment control unit 754 passes through the parameter calculation unit 743, the parameter line drawing unit 744, the intersection detection unit 745, and the overlap count measurement unit 746 so as to perform the processing in the leading candidate line drawing unit 747 and the subsequent steps. To control.

Next, the parameter fine adjustment control unit 754 controls the parameter space coarse-grain processing unit 742 and the powerful candidate line drawing unit 747 to perform processing in the second and subsequent iterations.
This is because, in the first stage (the first process), the parameter calculation unit 743, the parameter line drawing unit 744, and the intersection detection unit 745 overlap each other because the region is narrowed down from the entire parameter space by the Hough transform. Processing in the number measuring unit 746 is required.
On the other hand, in the second and subsequent processing stages, a real line is obtained once in the discretization width of a certain parameter space in the first stage (first process), and the real line is directly searched as a promising candidate line in the vicinity. Therefore, the parameters in the parameter calculation unit 743, the parameter line drawing unit 744, the intersection detection unit 745, and the overlap count measurement unit 746 become unnecessary.

  Here, the parameter space coarse-grain processing unit 742 of the present embodiment can also be referred to as “discretization width adjusting means”. The parameter fine adjustment control unit 754 can also be referred to as “second adjustment control means”.

  In this case, the discretization width adjusting means can adjust the discretization width of each coordinate axis of the parameter space portion. Further, the second adjustment control means determines the real line using the parameter space portion of the first discretization width obtained by increasing the discretization width by the discretization width adjustment means. The determined actual line is set as a new new candidate line, a new search is performed for a new candidate line peripheral region around the new candidate line, and the parameter of the second discretization width smaller than the first discretization width A new real line is newly determined using a space, the change of the discretization width and the determination of the new real line are repeatedly performed in stages, and adjustment control of the determination of the real line can be performed.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In the image processing method according to the present embodiment, first, based on the pixel value, the computer converts a candidate area and a background area into two areas, that is, a candidate area for a straight line or a curve to be detected and another area. Then, the image is divided into two regions (step S801) <binarization processing step>.

  Next, the computer performs processing for increasing the discretization width of the parameter space to a predetermined width (first discretization width) and coarsening the discretization width of the parameter space (step S802) <parameter space coarse-graining. Processing step>.

Subsequently, the computer determines whether the process is the first process or the second or later process (step S803) <process count determination process step>.
In this determination process, if the computer determines that the process is the first process, the computer performs the process from step S804. If the computer determines that the process is the second or subsequent process, the computer Performs the process from step S808.

  And a computer performs the process of step S804-step S813 which is the process sequence similar to the said 1st Embodiment.

Further, the computer determines whether or not the preset discretization width on the preset parameter space has been reached (step S814) <set discretization width determination processing step>.
If it is determined in this determination process that the computer has not reached the preset discretization width in the parameter space, the computer returns to step S801, and steps S801 to S803 and steps S808 to S813. Processing (second and subsequent binarization processing steps, second and subsequent parameter space coarse-graining processing steps, second and subsequent processing count determination processing steps, second and subsequent leading candidate line drawing processing steps, and second and subsequent processing steps The passage determination processing step, the second and subsequent weight calculation processing step, the second and subsequent number measurement processing step, the second and subsequent real line determination processing step, and the second and subsequent real line image generation processing step) are repeatedly executed.
On the other hand, if the computer determines that the preset discretization width on the parameter space has been reached, the computer ends the process.

  Here, in the second and subsequent processing stages, a real line is once obtained with a discretization width of a certain parameter space in the first stage (first process), and the real line is directly searched as a leading candidate line in the vicinity thereof. Since the parameters are refined, the processing from step S804 to step S807 becomes unnecessary.

  As described above, the computer increases the discretization width of the parameter space to a predetermined width to roughen the parameter space to obtain a real line, and once the real line is obtained, the parameter space is discretized more densely.

  Then, the computer re-determines a new real line in the vicinity of the real line once obtained in the subdivided parameter space, and the derivation of the new real line in the parameter space having a different discretization width is performed a predetermined number of times or It is repeatedly executed until the discretization width reaches a predetermined value.

  Here, the step of performing the repetitive control in steps S803 to S814 can also be referred to as a “second adjustment control step”. Step S802 can also be called a discretization width adjustment step. In the discretization width adjustment step, the discretization width of each coordinate axis of the parameter space portion is adjusted. In this case, in the “first adjustment control step”, the real line is determined using the parameter space portion of the first discretization width obtained by increasing the discretization width in the discretization width adjustment step. Later, the determined real line is set as a new new candidate line, a new search is performed for a new candidate line peripheral region around the new candidate line, and a second discretized width smaller than the first discretized width A new real line is newly determined using the parameter space, and the change of the discretization width and the determination of the new real line are repeated stepwise to perform adjustment control of the determination of the real line. .

  As described above, according to the present embodiment, while the same effect as the first embodiment is achieved, the position information of the detected line is improved by increasing the discretization width first and then gradually increasing the resolution. Since the parameter fine adjustment unit is provided, it is possible to detect the line with higher accuracy and lower calculation amount by sequentially increasing the definition of the parameter space.

  Other configurations, other steps, and operational effects thereof are the same as those in the first embodiment described above. In the above description, the operation content of each step described above and the components of each unit may be programmed and executed by a computer.

  In the present embodiment, for the sake of simplicity, the configuration example in which the parameter space coarse-grain processing unit and the parameter fine adjustment control unit are added to the image processing apparatus of the first embodiment has been described. When the parameter space coarse-grain processing unit and the parameter fine adjustment control unit are added to the image processing apparatus according to the embodiment, the parameter space coarse-graining is added to each of the first and second image processing apparatuses of the third embodiment. When adding a processing unit and parameter fine adjustment control unit, when adding a parameter space coarse-graining processing unit and parameter fine adjustment control unit to the image processing apparatus of the fourth embodiment, the fifth embodiment. When a parameter space coarse-grain processing unit and a parameter fine adjustment control unit are added to the image processing apparatus, the parameter space coarse-grain processing unit and the parameter fine adjustment control are added to the image processing apparatus according to the sixth embodiment. It may be a case to add.

(Modification of the seventh embodiment)
(About configuration)
FIG. 29 is a block diagram illustrating a configuration example in which a parameter space coarse-grain processing unit and a parameter fine adjustment control unit are added to the second embodiment.

  As shown in FIG. 29, the image processing apparatus 840 in the image processing system differs from the second embodiment in that all candidate line drawing units are provided in addition to the parameter space coarse-graining processing unit 842 and the parameter fine adjustment control unit 849. Instead of this, all candidate line leading candidate line drawing sections 843 are provided.

  The all candidate line leading candidate line drawing unit 843 first draws lines on the image for all parameter combinations on the parameter space of the first discretization width (the first time all candidate line drawing processing) Once the real line is obtained and the real line is obtained with the next discretization width (second discretization width), the line is drawn only for the combination of parameters in the parameter space near the found real line (2 From the first time on, the candidate line drawing process).

  In addition, the image processing apparatus 840 has the same configuration as that of the second embodiment, a binarization processing unit 841, a passage determination unit 844, a weight calculation unit 845, a number measuring unit 846, a real line A determination unit 847 and a real line image generation unit 848 are included.

  In the image processing apparatus 840 configured as described above, the parameter fine adjustment control unit 849 changes the discretization width of the parameter space in the parameter space coarse-graining processing unit 842 following the binarization processing unit 841. To control.

  After that, the parameter fine adjustment control unit 849 performs processing in all candidate line leading candidate line drawing units 843 and subsequent processing (processing in the first pass determination unit 844, processing in the first weight calculation unit 845, first time Control at the number measuring unit 846 of the first actual line determination unit 847).

  Next, the parameter fine adjustment control unit 849 similarly performs the processing in the parameter space coarse-graining processing unit 842, all candidate line leading candidate line drawing unit 843 and the subsequent processing (second and subsequent times) in the second and subsequent iterations. Processing in the passage determination unit 844, processing in the second and subsequent weight calculation unit 845, processing in the second and subsequent number measurement unit 846, processing in the second and subsequent actual line determination unit 847) are performed. To control.

  Here, in the first stage (the first process), all candidate line leading candidate line drawing unit 843 draws lines on the image for all parameter combinations in the parameter space. I do.

  On the other hand, in the second and subsequent processing stages, in the first stage (first processing), the actual line is once obtained with a certain discretization width (first discretization width) in the parameter space, and the next discretization width is obtained. When calculating a real line with (second discretization width), draw a line only for the combination of parameters when the real line found in the parameter space is searched as a promising candidate line in the vicinity, and the parameter is precisely To perform leading candidate line drawing processing.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In the image processing method according to the present embodiment, first, based on the pixel value, the computer converts a candidate area and a background area into two areas, that is, a candidate area for a straight line or a curve to be detected and another area. Then, the image is divided into two regions (step S901) <binarization processing step>.

  Next, the computer performs a process of increasing the discretization width of the parameter space to a predetermined width (first discretization width) and coarsening the discretization width of the parameter space (step S902) <parameter space coarse-graining. Processing step>.

Subsequently, the computer performs all candidate line leading candidate line drawing processing step in step 903.
Here, it is determined whether the process is the first process or the second and subsequent processes <process count determination process step>. In this determination process, if the computer determines that this is the first process, the computer draws lines on the image for all parameter combinations in the parameter space. In the line drawing process, all candidate line drawing processes are performed.

  On the other hand, when the computer determines that the process is the second or later process, the computer temporarily uses a discretization width (first discretization width) having a parameter space in the first stage (first process). Since the real line is obtained and the real line is obtained with the next discretization width (second discretization width), in the all candidate line leading candidate line drawing process, the found real line is directly used as the leading candidate line in the parameter space. A line is drawn only for the combination of parameters when searching in the vicinity, and the leading candidate line drawing process is performed by refining the parameters.

  Then, the computer performs the processing from step S904 to step S908, which is the same processing procedure as in the second embodiment.

Further, the computer determines whether or not the preset discretization width on the preset parameter space has been reached (step S909) <set discretization width determination processing step>.
If it is determined in this determination process that the computer has not reached the preset discretization width in the parameter space, the computer returns to step S901 and performs the processes in steps S901 to S908 (second and subsequent times). Binarization processing step, second and subsequent parameter space coarse-graining processing step, second and subsequent all candidate line leading candidate line drawing processing step, second and subsequent passage determination processing step, and second and subsequent weight calculation processing step The second and subsequent number counting processing steps, the second and subsequent real line determination processing steps, and the second and subsequent real line image generation processing steps) are repeatedly executed.
On the other hand, if the computer determines that the preset discretization width on the parameter space has been reached, the computer ends the process.

  Here, in the second and subsequent processing stages, a real line is once obtained with a discretization width of a certain parameter space in the first stage (first process), and the real line is directly searched as a leading candidate line in the vicinity thereof. The parameters will be refined.

  As described above, the computer increases the discretization width of the parameter space to a predetermined width to roughen the parameter space to obtain a real line, and once the real line is obtained, the parameter space is discretized more densely.

  Then, the computer re-determines a new real line in the vicinity of the real line once obtained in the subdivided parameter space, and the derivation of the new real line in the parameter space having a different discretization width is performed a predetermined number of times or It is repeatedly executed until the discretization width reaches a predetermined value.

[Eighth Embodiment]
Next, an eighth embodiment according to the present invention will be described with reference to FIG. In the following, description of the substantially similar configuration of the first embodiment will be omitted, and only different parts will be described. FIG. 31 is a block diagram showing an example of the eighth embodiment of the image processing apparatus in the image processing system of the present invention.

  In the present embodiment, the parameter line drawing unit with a pixel value weight configured instead of the parameter line drawing unit of the image processing apparatus of the first embodiment, and the image processing apparatus of the first embodiment. A pixel value weighted overlap number measuring unit configured in place of the overlap number measuring unit, and a pixel value weighted weight calculating unit configured in place of the weight calculating unit of the image processing apparatus according to the first embodiment. A configured example is disclosed.

(About configuration)
Specifically, the image processing apparatus 940 in the image processing system of the eighth embodiment includes a pixel value weighted parameter line drawing unit 943 and a pixel value weighted overlap count measuring unit 945 as shown in FIG. A pixel value weighted weight calculation unit 948, a binarization processing unit 941, a parameter calculation unit 942, an intersection detection unit 944, and a powerful candidate line drawing having the same configuration as that of the first embodiment. Unit 946, passage determination unit 947, number measurement unit 949, real line determination unit 951, real line image generation unit 952, pixel value weight calculation unit 954 for calculating pixel value weights, and execution procedures of these units And a module control unit 953 for controlling.

The pixel value weighted parameter line drawing unit 943 is a means (pixel value weight calculating unit 954) that predetermines a weight based on the pixel value of the image before binarization when drawing the parameter line on the parameter space. The pixel value weight is calculated to be drawn in consideration of the pixel value weight.
For example, the pixel value weight corresponding to the pixel is given to the number of elements of which the coordinates of the parameter line and the parameter space portion match and rendered.

  The pixel value weighted overlap count metric unit 945 is a coordinate whose number of parameter lines overlaps at the intersection and the sum of the pixel value weights is larger than a threshold given in advance from the outside, and other within a predetermined neighborhood. The combination of the parameters indicated by the coordinates having no portion exceeding the number of times of overlap is set as a promising candidate for the parameter combination of the straight line or the curve to be detected.

The pixel value weighted weight calculation unit 948 obtains the tangent line of the leading candidate line at the target pixel, and obtains the absolute value of the reciprocal of the cosine (cos θ) of the angle formed between the tangent line and the x axis. If this absolute value is less than or equal to the square root of 2, the absolute value of the reciprocal of the cosine of this angle is the slope weight, and otherwise, the absolute value of the reciprocal of sine (sin θ) is the slope weight.
The inclination weight is calculated for each passing pixel of the leading candidate line. In addition, the inclination weight is calculated for all passing pixels of all possible candidate lines, and is stored in, for example, the parameter space storage unit 122.
The pixel value weighted weight calculation unit 948 then multiplies the inclination weight by a value calculated based on the pixel value of the image before binarization to obtain a new inclination weight (pixel value weighted inclination weight).

  As the pixel value weight in each part, for example, the pixel value may be used as it is, or the pixel value may be logarithmically converted or may be converted by a function given in advance.

  Here, the pixel value weight calculation unit 954 of the present embodiment can also be referred to as “pixel value weight calculation means”. The pixel value weighted weight calculating unit 948 and the number measuring unit 949 of the present embodiment can also constitute a “candidate line identification index calculating unit”. Further, the “parameter space drawing unit” may be configured by the pixel value weighted parameter line drawing unit 943 and the parameter calculation unit 942. Further, the intersection detection unit 944 and the pixel value weighted overlap number measurement unit 945 may constitute a “selection unit”.

  The pixel value weight calculating means can calculate a specific pixel value weight based on a pixel value of the image before extracting the candidate area. Further, in this case, the candidate line identification index calculating means calculates a pixel value weighted inclination weight for each passing pixel based on the pixel value weight and the inclination weight, and the pixel value for each candidate line. A total value of weighted gradient weights can be calculated.

  The parameter space drawing unit draws the pixel value weight corresponding to the pixel, the parameter space graphic element, and the parameter space part when drawing the parameter space graphic element corresponding to the pixel in the parameter space part. It is possible to draw by assigning to the number of elements having the same coordinates. In this case, the selecting means is configured to determine whether the parameter space portion of the parameter space portion is based on the total number of the pixel value weights at the intersection where each of the parameter space graphic elements intersects and a preset pixel value weight total number threshold value. Possible candidate areas in the parameter space of the intersection and the surrounding area of the intersection can be selected.

  Further, when the pixel value weight calculating unit calculates a pixel value weight by logarithmically converting the pixel value, the candidate line identification index calculating unit performs logarithmic conversion on each candidate line. The total value of the pixel value weighted gradient weights can be calculated.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In the image processing method according to the present embodiment, first, the computer performs a process of calculating the pixel value weight in order to predetermine the weight based on the pixel value of the image before binarization (step S1000a) <pixel Value weight calculation processing step>.

  Next, the computer divides the image into two regions based on the pixel values, with the two regions of the detection target line and curve candidates and the other regions as a candidate region and a background region, respectively. (Step S1001) <binarization processing step>.

  Then, the computer performs a parameter calculation process in step S1002, which is the same processing procedure as in the first embodiment.

  Next, when the computer draws the parameter line in the parameter space, the pixel value weight is calculated by calculating a weight based on the pixel value of the image before binarization and taking this pixel value weight into consideration. Then, drawing is performed (step S1003) <Pixel value weighted parameter line drawing processing step>.

  Subsequently, the computer performs an intersection detection process in step S1004, which is the same processing procedure as in the first embodiment.

  Further, the computer has coordinates where the number of parameter lines overlap each other at the intersection and the sum of the pixel value weights is larger than a threshold given in advance from the outside, and there is no other overlapping portion within a predetermined neighborhood. A process is performed in which the combination of parameters indicated by the coordinates is a promising candidate for the parameter combination of a straight line or a curve to be detected (step S1005) <pixel value weighted overlap count measurement process step>.

  Next, the computer performs the processing of steps S1006 to S1007, which is the same processing procedure as in the first embodiment.

  Then, the computer performs processing for newly obtaining a gradient weight by multiplying the gradient weight by a value calculated based on the pixel value of the image before binarization (step S1008) <pixel value weighted weight calculation processing Step>.

  Next, the computer performs the processing from step S1009 to step S1011 which is the same processing procedure as in the first embodiment.

  Here, step S1000a of the present embodiment can also be referred to as a “pixel value weight calculation step”. Moreover, it can also be called "candidate line identification parameter | index calculation step" by step S1008 and step S1009 of this Embodiment. Furthermore, it can also be referred to as a “parameter space drawing step” by steps S1002 and S1003 of the present embodiment. Moreover, it can also be called "selection step" by step S1004 and step S1005 of this Embodiment.

  In the pixel value weight calculation step, a specific pixel value weight is calculated based on the pixel value of the image before extracting the candidate area. In this case, in the candidate line identification index calculation step, an inclination weight with a pixel value weight is calculated for each passing pixel based on the pixel value weight and the inclination weight, and the pixel value weight is added to each candidate line. The total value of the inclination weight can be calculated.

In the parameter space portion drawing step, when the parameter space graphic element corresponding to the pixel is drawn in the parameter space portion, the pixel value weight corresponding to the pixel is set to the parameter space graphic element and the parameter space portion. It is possible to draw by assigning to the number of elements having the same coordinates.
In the selection step, based on the total number of pixel value weights at the intersection where each of the parameter space graphic elements intersects and a preset pixel value weight total number threshold, It is possible to select potential candidate areas in the parameter space around the intersection.

  In the pixel value weight calculation step, the pixel value weight can be calculated by logarithmically converting the pixel value. In this case, in the candidate line identification index calculating step, a total value of the pixel value weighted inclination weights logarithmically converted for each of the candidate lines can be calculated.

  As described above, according to the present embodiment, the same effect as the first embodiment can be obtained, but by using the pixel value weight, a line that is clearly visible but has a short length is displayed. Can be detected.

  Other configurations, other steps, and operational effects thereof are the same as those in the first embodiment described above. In the above description, the operation content of each step described above and the components of each unit may be programmed and executed by a computer.

  In this embodiment, for the sake of simplicity, a configuration example in which the pixel value weight is considered in the image processing apparatus according to the first embodiment (a pixel value weighted parameter line drawing unit, a pixel value weighted overlap count measuring unit) An example in which the pixel value weighted weight calculation unit is configured) has been described, but when the pixel value weight is considered in the image processing apparatus of the second embodiment (a pixel value weighted parameter line drawing unit, pixel value weight) In the case where an appending overlap count measurement unit and a pixel value weighted weight calculation unit are configured), in the case where each of the first and second image processing apparatuses of the third embodiment considers a pixel value weight (pixel value weight) A parameter line drawing unit, a pixel value weighted overlap count measurement unit, and a pixel value weighted weight calculation unit), when the pixel value weight is considered in the image processing apparatus of the fourth embodiment (pixel value) Weighted parameter line drawing When the pixel value weighted overlap count measuring unit and the pixel value weighted weight calculating unit are configured), when the pixel value weight is considered in the image processing apparatus according to the fifth embodiment (pixel value weighted parameter line drawing) A pixel value weighted overlap count measuring unit, and a pixel value weighted weight calculating unit), and a pixel value weighted parameter line in the image processing apparatus according to the sixth embodiment (pixel value weighted parameter line) When a drawing unit, a pixel value weighted overlap count measurement unit, and a pixel value weighted weight calculation unit are configured), when the pixel value weight is considered in the image processing apparatus according to the seventh embodiment (pixel value weighted parameter) A line drawing unit, a pixel value weighted overlap count measuring unit, and a pixel value weighted weight calculating unit).

(First Modification of Eighth Embodiment)
(About configuration)
FIG. 33 is a block diagram illustrating a configuration example when pixel value weights are considered in the image processing apparatus according to the second embodiment (when a pixel value weighted weight calculation unit is configured).

  As shown in FIG. 33, the image processing apparatus 1040 in the image processing system has a configuration in which a pixel value weighted weight calculation unit 1044 is provided instead of the weight calculation unit, unlike the second embodiment.

The pixel value weighted weight calculation unit 1044 obtains the tangent of the candidate line at the target pixel, and obtains the absolute value of the reciprocal of the cosine (cos θ) of the angle formed between the tangent and the x axis. If this absolute value is less than or equal to the square root of 2, the absolute value of the reciprocal of the cosine of this angle is the slope weight, and otherwise, the absolute value of the reciprocal of sine (sin θ) is the slope weight.
The inclination weight is calculated for each pixel through which the candidate line passes. In addition, the inclination weight is calculated for all candidate lines and is held in, for example, the parameter space storage unit 122.
Then, the pixel value weighted weight calculation unit 1044 multiplies the inclination weight by a value calculated based on the pixel value of the image before binarization to obtain a new inclination weight (pixel value weighted inclination weight).

  In addition, the image processing apparatus 1040 has the same configuration as that of the second embodiment, a binarization processing unit 1041, a passage determination unit 1044, a weight calculation unit 1045, a number measuring unit 1046, a real line A determination unit 1047 and a real line image generation unit 1048 are included.

(About processing procedure)
Next, the processing of each unit in the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG.

  In this image processing method, first, based on pixel values, a computer uses two regions, a candidate region for a straight line and a curve to be detected, and another region as a candidate region and a background region, respectively. Divide into two areas (step S1101) <binarization processing step>.

  Next, the computer performs the processing from step S1002 to step S1103, which is the same processing procedure as in the second embodiment.

Subsequently, the computer obtains the tangent of the candidate line at the target pixel, and obtains the absolute value of the reciprocal of the cosine (cos θ) of the angle formed by the tangent and the x axis. If this absolute value is less than or equal to the square root of 2, the absolute value of the reciprocal of the cosine of this angle is the slope weight, and otherwise, the absolute value of the reciprocal of sine (sin θ) is the slope weight.
The inclination weight is calculated for each pixel through which the candidate line passes. In addition, the inclination weight is calculated for all candidate lines.
Then, the computer further multiplies the inclination weight by a value calculated based on the pixel value of the image before binarization to obtain a new inclination weight (inclination weight with pixel value weight) (step S1104) <pixel value weight Added weight calculation processing step>.

  According to the first modified example of the embodiment, although the same effect as that of the second embodiment is achieved, the pixel value weight is clearly used, but the length is clearly shown. Short lines can be detected.

(Second Modification of Eighth Embodiment)
(About configuration)
FIG. 35 shows a case where a pixel value weight is considered in the switching control device of the third embodiment (a pixel value weighted parameter line drawing calculation amount estimation unit, a pixel value weighted overlap count metric calculation amount estimation unit, a pixel value, It is a block diagram showing a configuration example of a weighted weight calculation calculation amount estimation unit and an all pixel value weighted weight calculation calculation amount estimation unit).

  As shown in FIG. 35, the switching control device 1170 in the image processing system is different from the third embodiment in that a parameter line drawing calculation amount estimation unit with a pixel value weight is used instead of the parameter line drawing calculation amount estimation unit. 1173, a pixel value weighted overlap count metric calculation amount estimation unit 1175 is provided instead of the overlap count metric calculation amount estimation unit, and a pixel value weighted weight calculation calculation amount estimation unit 1178 is provided instead of the weight calculation calculation amount estimation unit. A configuration example is disclosed in which an all-pixel value weighted weight calculation calculation amount estimation unit 1184 is provided instead of the total weight calculation calculation amount estimation unit.

  As shown in FIG. 35, the switching control device 1170 includes a pixel value weighted parameter line drawing calculation amount estimation unit 1173, a pixel value weighted overlap count metric calculation amount estimation unit 1175, and a pixel value weighted weight calculation calculation amount estimation. Unit 1178 and an all-pixel-value-weighted weight calculation calculation amount estimation unit 1184.

  Furthermore, as shown in FIG. 35, the switching control device 1170 has the same configuration as that of the third embodiment, a binarization processing unit 1171, a parameter calculation calculation amount estimation unit 1172, and an intersection detection calculation amount. Estimator 1274, leading candidate line drawing calculation amount estimation unit 1176, passage determination calculation amount estimation unit 1177, number metric calculation amount estimation unit 1179, real line determination calculation amount estimation unit 1186, and all candidate line drawing calculation amounts An estimation unit 1182, an all-pass determination calculation amount estimation unit 1183, a total piece count calculation amount estimation unit 1185, an all real line determination calculation amount estimation unit 1186, a method selection unit 1187, and the execution procedure of these units are controlled. A module control unit 1188.

  The pixel value weighted parameter line drawing calculation amount estimation unit 1173 estimates the calculation amount required for the execution of the pixel value weighted parameter line drawing unit.

  The pixel value weighted overlap count metric calculation amount estimation unit 1175 estimates the calculation amount required for the pixel value weighted overlap count metric calculation unit.

  The pixel value weighted weight calculation calculation amount estimation unit 1178 estimates the calculation amount required to execute the pixel value weighted weight calculation unit.

The all pixel value weighted weight calculation calculation amount estimation unit 1184 estimates the calculation amount required for the execution of the all pixel value weighted weight calculation unit.
That is, the all-pixel-value-weighted weight calculation calculation amount estimation unit 1184 determines that the absolute value of the reciprocal of the cosine of the angle formed by the tangent line and the x axis on the image is less than the square root of 2 according to the tangent slope of all candidate lines. The absolute value of the reciprocal of the cosine of this angle is used as the inclination weight, and the absolute value of the reciprocal of the sine is used as the inclination weight, and the amount of calculation required to obtain the inclination weight for each passing candidate line is estimated. The weight calculation calculation amount.

(About processing procedure)
Next, the processing of each unit in the switching control device of the image processing system having the above-described configuration can be realized as a method, and various processing procedures as the image processing method will be described with reference to FIG. .

  In this image processing method, first, steps S1001 to S1011 shown in FIG. 32, which are the same processes as those in the eighth embodiment, are performed. Next, steps S1101 to S1107 shown in FIG. 34, which are the same processes as those of the first modification of the eighth embodiment, are performed.

  Here, in particular, in step 1003, when the computer draws the parameter line in the parameter space, the computer calculates the weight based on the pixel value of the image before binarization (pixel value weight calculation processing). Then, pixel value weighting is performed and drawing is performed in consideration of the pixel value weighting <parameter value weighted parameter line drawing processing step>.

  Further, in step 1005, the computer is such that the number of overlapping parameter lines and the sum of the pixel value weights at the intersection are coordinates larger than a threshold given from the outside in advance, and the other overlaps within a predetermined neighborhood. <Pixel value weighted overlap count measurement processing step> A combination of parameters indicated by coordinates having no portion exceeding the number of times is set as a potential candidate for a parameter combination of a straight line or a curve to be detected.

  Furthermore, in step 1008, the computer sets a new inclination weight by multiplying the inclination weight for the leading candidate line by a value calculated based on the pixel value of the image before binarization <first pixel value weight Added weight calculation processing step>.

  Furthermore, in step 1104, the slope weight for the candidate line is further multiplied by a value calculated based on the pixel value of the image before binarization to obtain a new slope weight <second pixel value weighted weight calculation process Step>.

  Next, after the processing in steps S1001 to S1011 shown in FIG. 32 and the processing in steps S1101 to S1107 shown in FIG. 34, the computer determines areas and candidates for straight lines or curves that are detection targets based on the pixel values. The two regions other than the region are the candidate region and the background region, respectively, and the image is divided into two regions, the ratio of the candidate regions on the binarized image is obtained, and the discretization width is set for each axis of the parameter space. The amount of calculation for each process is estimated (step S1201) <binarization process step>.

  Subsequently, the computer performs a parameter calculation calculation amount estimation process in step S1202, which is the same processing procedure as in the third embodiment.

  Then, the computer estimates the amount of calculation required to execute the pixel value weighted parameter line drawing process in step 1003 (step S1203) <pixel value weighted parameter line drawing calculation amount estimation processing step>.

  Subsequently, the computer performs an intersection detection calculation amount estimation process in step S1202, which is the same processing procedure as in the third embodiment.

  Further, the computer estimates the amount of calculation required to execute the pixel value weighted overlap count metric processing in step 1005 (step S1205) <pixel value weighted overlap count metric calculation amount estimation processing step>.

  Then, the computer performs a calculation amount estimation process in steps S1206 to S1207, which is the same processing procedure as in the third embodiment.

  Further, the computer estimates the amount of calculation required to execute the first pixel value weighted weight calculation processing of step 1008 for the probable candidate line (step S1205) <pixel value weighted weight calculation calculation amount estimation processing step> .

  Next, the computer performs a calculation amount estimation process in steps S1209 to S1212 which is the same processing procedure as that of the third embodiment.

  Subsequently, the computer estimates a calculation amount required to execute the second pixel value weighted weight calculation process of step 1104 for the candidate line (step S1213) <all pixel value weighted weight calculation calculation amount estimation processing step. >.

  Furthermore, the computer performs a calculation amount estimation process in steps S1214 to S1215, which is the same processing procedure as in the third embodiment.

Then, the computer calculates the parameter calculation calculation amount, the pixel value weighted parameter line drawing calculation amount, the intersection detection calculation amount, the pixel value weighted overlap count metric calculation amount, the leading candidate line drawing calculation amount, the passage determination calculation amount, and the pixel value. The sum of the weighted weight calculation amount, the number metric calculation amount, and the actual line determination calculation amount is defined as a pixel value weighted parameter space calculation amount (first calculation amount).
In addition, the computer calculates the sum of all candidate line drawing calculation amount, all pass determination calculation amount, all pixel value weighted weight calculation calculation amount, all number metric calculation amount, and all real line determination calculation amount as a pixel value weighted image. Let it be a spatial calculation amount (second calculation amount).
In addition, the computer compares the parameter space calculation amount with the image space calculation amount, and if the pixel value weighted parameter space calculation amount is smaller than the pixel value weighted image space calculation amount, A real line image is generated by the processing procedure of steps S1001 to S1011 of the embodiment.
On the other hand, when the pixel value weighted image space calculation amount is smaller than the pixel value weighted parameter space calculation amount, the computer performs the processing procedure of steps S1101 to S1107 of the first modification of the eighth embodiment. The real line image by is generated.
On the other hand, when the pixel value weighted image space calculation amount is equal to the pixel value weighted parameter space calculation amount, a real line image is generated by one of the processing procedures based on a predetermined rule (step S1216) < Method selection processing step>.

  According to the second modified example of such an embodiment, although the same effect as that of the third embodiment is achieved, the pixel value weight is clearly used, but the length is clearly shown. Short lines can be detected.

[Various modifications]
Also, although the apparatus and method according to the present invention have been described according to some specific embodiments thereof, the embodiments described in the text of the present invention can be used without departing from the spirit and scope of the present invention. Various modifications are possible.
For example, an image processing system including the image processing apparatuses according to the first to eighth embodiments and a switching control apparatus that controls the switching of the image processing apparatuses may be configured.

  Further, the number, position, shape, and the like of the constituent members are not limited to the above-described embodiment, and can be set to a number, position, shape, and the like that are suitable for implementing the present invention. That is, in the third embodiment, the case where there are two image processing devices, one input device, one output device, and one storage device is shown. It is not limited.

Furthermore, as an example of the Hough transform, an example is given in which a graphic element (point) in a two-dimensional orthogonal coordinate system is converted into a graphic element (parameter line) on a two-dimensional polar coordinate system of ρ-θ, which is an example of a parameter space. , It may be converted into other various parameter spaces.
For example, assuming that the coordinates of a point on a circle having a center coordinate (a, b) radius r in the two-dimensional orthogonal coordinate system are (x _i , y _i ), all circles that can pass through this pixel have a, b, r as It is represented by a circle of (ax _i ) ² + (b−y _i ) ² = r ² in the parameter space as an axis. The point in the two-dimensional orthogonal coordinate system may be converted into a graphic element on a three-dimensional coordinate system of a, b, and r, which is an example of a parameter space. In this case, a point in the two-dimensional orthogonal coordinate system corresponds to one surface on the three-dimensional coordinate system (parameter space) of a, b, and r. When there are a large number of points, a large number of curved surfaces are obtained in the parameter space. A point shared by each of these curved surfaces is detected.

  Furthermore, it is not limited to a circle that can pass through pixels, but may be an ellipse.

The lines to be detected by the Hough transform are generally known in advance, such as straight lines and circles. However, from the function part as in the generalized Hough transform, which is an extension of the Hough transform. The present invention can be applied even when a figure having an arbitrary shape is not detected.
Further, as the parameter space, an extended parameter line may be generated by arbitrarily distorting the ρ-θ parameter plane (in the same phase as the parameter line).
Further, a parameter space of a and b of a line y = ax + b that can pass through the pixel may be used.

  In this way, image processing that converts the coordinates of each pixel on the target figure expressed in the Cartesian coordinate system to a coordinate system in the parameter space, and recognizes the line figure on the image by the shape on the coordinates in the parameter space When a feature point in the image space is given for the entire device, it is assumed that all the graphic elements including the point exist, and the points in the parameter space representing the graphic element are voted, and each point in the parameter space is In general, image processing devices that determine whether or not there is a graphic element represented by each point by examining how many feature points received the vote, and using the coefficient of the equation representing the line to be detected as a parameter For each line, find all the parameter combinations for all the lines that can pass through that point, and plot all the parameter combinations in the parameter space with this parameter as the axis. And a parameter line draw parameter line for each point in the image, a point that overlaps many more parameters lines are applicable to an image processing device which is a combination of parameters to be detected lines.

  Further, the total number of inclination weights of the target pixel may be output by the number measuring unit. In this case, the Euclidean distance for the leading candidate line composed of each pixel can also be calculated from the total value of the inclination weights of the target pixel.

  In addition, for several points on the outline of an irregular figure, a template of an arbitrarily shaped figure that defines the reference point of the point, the displacement from the reference direction and the tangential direction is constructed, and the edge strength of a point in the figure space If the edge direction is detected and the point is considered to be on the edge, it is considered that the template passes through the point by a similar transformation (translation, scaling, rotation). At this time, the parameter space may be represented by a similarity transformation parameter (a two-dimensional translation space, a rotation scalar, and a four-dimensional parameter space using a scaling scalar).

  Further, the image processing apparatus 40 operates by program control, and if it has a network-related function, the information processing apparatus having a wireless / wired communication function or a similar computer, etc. Any computer can be used, mobile or fixed.

(program)
Further, the software program of the present invention that realizes the functions of the above-described embodiments is a program corresponding to the processing unit (processing means), functions, etc. shown in the various block diagrams in each of the above-described embodiments, Each processing program processed in a program corresponding to a processing procedure, processing means, function, etc. shown in a flowchart etc., a method (step) generally described in this specification, processing explained, data The whole or each part is included.

  Specifically, the program of the present invention represents a line that can pass through a pixel in an image by an equation including a specific parameter, and uses a parameter space portion having the parameter of the equation as a coordinate axis to make a line from the image. This is intended for a computer executable process.

  This program is a candidate line drawing control function (for example, in FIG. 1) that performs control for drawing each candidate line in the image corresponding to the coordinate value based on the coordinate value taken by the combination of the parameters in the parameter space portion. The function corresponding to the reference numeral 40a shown) can be included.

  This program also has a candidate line peripheral area determination function (for example, the reference numeral 47 shown in FIG. 1) that determines the candidate pixels and passing pixels through which the candidate line peripheral areas around the candidate lines pass. Corresponding functions). Further, the program calculates an inclination weight that is an index for identifying the candidate line and indicates an inclination of a tangent line of the candidate line for each passing pixel, and calculates a total value of the inclination weights for each candidate line. A candidate line identification index calculation function (for example, a function corresponding to the reference numeral 40d shown in FIG. 1) can be included.

  Furthermore, the program is configured so that one parameter space unit corresponding to one candidate line is based on the total value of each candidate line calculated by the candidate line identification index calculation function. In the parameter value peripheral area around the coordinate value of the parameter combination, search for the presence or absence of coordinate values of another parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line. When there is no total value of the other candidate lines that exceeds the total value of the one candidate line in the parameter value peripheral region, control is performed to determine one candidate line as a real line. A line determination control function (for example, a function corresponding to reference numeral 51 shown in FIG. 1 or the like) can be included.

  Further, the candidate line drawing control function intersects the parameter space drawing element with a parameter space drawing function that draws a parameter space drawing element corresponding to the pixel for each pixel, and the parameter space drawing element. The intersection corresponding to the candidate line in the image based on the number of intersections in the intersection, the number of intersections in the other intersections around the intersection, and a preset intersection count threshold. A selection function for selecting potential candidates for coordinate values of the parameter combinations in the parameter space portion, and for each coordinate value in the potential candidate area around the coordinate values of the potential candidates selected by the selection function, And a candidate line drawing function for drawing a corresponding candidate line in the image.

  In the program, the real line is determined by an image reduction processing function for reducing the image and a first reduced image obtained by reducing the image at a first magnification by the image reduction processing unit. Later, the determined actual line is used as a new new candidate line, and a new search is performed for the area around the new candidate line around the new candidate line, and reduced at a second magnification higher than the first magnification. A new real line is newly determined using the parameter space in the second reduced image, the reduction and the new real line are repeatedly determined in stages, and adjustment control for determining the real line is performed. 1 adjustment control function.

  Still further, in the program, a discretization width adjustment function for adjusting a discretization width of each coordinate axis of the parameter space portion and a first discretization width obtained by increasing the discretization width by the discretization width adjustment means. After determining the real line using the parameter space portion, the determined real line is set as a new new candidate line, and a new search is performed for a new candidate line peripheral region around the new candidate line, A new real line is newly determined using the parameter space having a second discretization width smaller than the first discretization width, and the change of the discretization width and the determination of the new real line are repeated stepwise. And a second adjustment control function for performing adjustment control for determining the real line.

  Further, the program may further include a pixel value weight calculation function for calculating a specific pixel value weight based on a pixel value of the image before extracting the candidate area. In this case, the candidate line identification index calculation function calculates a pixel value weighted inclination weight for each passing pixel based on the pixel value weight and the inclination weight, and the pixel value weighted for each candidate line. The total value of the inclination weight can be calculated.

  In the parameter space drawing function, when the parameter space graphic element corresponding to the pixel is drawn in the parameter space part, the pixel value weight corresponding to the pixel is set to the parameter space graphic element and the parameter. It is possible to draw by assigning to the number of elements having the same coordinate in the space portion. In this case, in the selection function, based on the total number of the pixel value weights at the intersection where each of the parameter space graphic elements intersects and a preset pixel value weight total number threshold value, Possible candidate areas in the parameter space of the intersection and the surrounding area of the intersection can be selected.

  Further, the pixel value weight calculation function can calculate a pixel value weight by logarithmically converting the pixel value. In this case, the candidate line identification index calculation function can calculate the total value of the gradient weights with the pixel value weights for each of the candidate lines.

  Further, in the program, a first process in the case of performing drawing processing in the parameter space portion by the parameter space portion drawing function and drawing processing of each selected candidate line in the image by the candidate line drawing means. Based on the first calculation amount based on the calculation pattern and the second calculation amount based on the second calculation pattern when only drawing processing of each candidate line in the image is performed. A calculation pattern selection control function for performing control to select a calculation pattern can be further provided.

  Further, the program may further include a candidate area extracting function for extracting a candidate area from which the line can be extracted from the image. In this case, the candidate line drawing control function can draw each candidate line in the candidate area. In the real line determination control function, a real line can be determined from each of the candidate lines in the candidate area.

  The program may further include an expansion / contraction process function for performing one or both of the processes for expanding and contracting the candidate area a specific number of times. In this case, the candidate line drawing control function can draw the candidate line based on the image of the candidate area processed by the expansion / contraction process function.

Furthermore, the program can further have a smoothing processing function for smoothing the image before extracting the candidate region. In this case, the candidate line drawing control function can draw the candidate line based on the image of the candidate area smoothed by the smoothing processing function.

  As a method of supplying the program, it is also possible to provide the program from an external device that is communicably connected to the computer via an electric communication line (whether wired or wireless).

  According to the program of the present invention, if the program is read into a computer (CPU) from a storage medium such as a ROM storing the program and executed, or the control program is downloaded to the computer via communication means. If it is executed after the above, the apparatus according to the present invention described above can be realized relatively easily. When the software of the apparatus is embodied as an embodiment of the idea of the invention, it naturally exists and is used on a storage medium storing such software.

Moreover, the program is the same without any question about the copying stage of the primary copy product, the secondary copy product, etc. If the program is supplied using a communication line, the communication line becomes a transmission medium and the present invention is used.

  Moreover, the structure which recorded the above-mentioned program on the information recording medium may be sufficient.

In the above-described embodiment, the client server system or peer-to-peer (Peer) in which terminals form a network and exchange data with each other.
You may comprise as a system by To Peer) communication.

  Further, the image processing apparatus on which the above-described program is mounted is not limited to a personal computer, for example, but various servers, EWS (engineering workstation), medium-sized computer, mainframe, portable information terminal, various mobile terminals, PDA, It may be configured to be usable from a mobile phone, a wearable information terminal, various TVs / DVD recorders / various audio devices (such as portable), home appliances equipped with various information communication functions, game devices having network functions, and the like. . Or what was improved as an application displayed on these apparatuses can also be included in the scope of the present invention.

  The program may be a program for realizing a part of the above-described functions. Further, a program that can realize the above-described functions in combination with a program already recorded in a computer system, a so-called difference file ( Difference program).

  Further, in the present specification, the steps shown in the flowchart include processes that are executed in parallel or individually even if they are not necessarily processed in time series, as well as processes that are executed in time series according to the described procedure. It is a waste. In the implementation, the order in which the program procedures (steps) are executed can be changed. Further, certain procedures (steps) described herein can be implemented, removed, added, or rearranged as a combined procedure (step) as needed for implementation.

  Furthermore, the functions of the program such as each means of the apparatus, each function, and the procedure function of each step may be achieved by dedicated hardware (for example, a dedicated semiconductor circuit). Some of these functions may be processed by hardware, and other functions among all functions may be processed by software. In the case of dedicated hardware, each unit may be formed by an integrated circuit such as an LSI. These may be individually made into one chip, or may be made into one chip so as to include a part or all of them. Further, the LSI may include other functional blocks such as a streaming engine. Further, the method of circuit integration is not limited to LSI, and implementation with a dedicated circuit or a general-purpose processor is also possible. Further, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology.

  In addition, the communication structure between the image processing apparatus and the output apparatus, the communication structure between the image processing apparatus and the input apparatus, and the communication structure between the image processing apparatus and the storage device are formed in one or both of them. The interface type may be, for example, a parallel interface, USB interface, IEEE 1394, a network such as a LAN or WAN, or the like, or any interface that will be developed in the future.

  Furthermore, when performing a process of extracting a line from the image using the Hough transform and the inverse Hough transform, or the inverse Hough transform, after the inverse Hough transform, the passing pixels in the peripheral area of the candidate line Control for determining the Euclidean distance of each pixel of the candidate line based on the inclination of the tangent line in the pixels constituting the candidate line, and determining the candidate line having the maximum Euclidean distance among the candidate lines as the actual line It is not necessarily limited to a substantial apparatus, and it can be easily understood that this technique also functions as a method. For this reason, the invention relating to the method is not necessarily limited to a substantial apparatus, and there is no difference that the method is also effective. In this case, a device, a device, or the like can be included as an example for realizing the method.

  By the way, such an apparatus may exist independently, or may be used in a state where it is incorporated in a certain device (for example, a processor). Is included. Therefore, it can be changed as appropriate, such as software or hardware. When the software of the apparatus is embodied as an embodiment of the idea of the invention, it naturally exists on the storage medium storing the software and is used.

  Furthermore, it may be a case where a part is software and a part is realized by hardware, and a part is stored on a storage medium and is read as needed. It may be as a thing.

  The scope of the invention is not limited to the illustrated example. Furthermore, examples include combinations of the above-described embodiments, or any of them and any of the modifications. In addition, an embodiment based on a configuration in which some of the configuration requirements are deleted from all the configuration requirements shown in the embodiment, and a technical scope based on the configuration may be an invention. The above description is merely an example of the embodiment, is illustrative, not limiting, and can be modified and / or changed as appropriate.

  The present invention can be applied to all image processing apparatuses capable of recognizing line figures on an image.

1 is a block diagram showing an example of the overall configuration of an image processing system according to a first embodiment of the present invention. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the 1st Embodiment of this invention. It is a block diagram which shows an example of the whole structure of the image processing system by the 2nd Embodiment of this invention. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is explanatory drawing for demonstrating the content processed with the image processing apparatus in the image processing system of FIG. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the 2nd Embodiment of this invention. It is a block diagram which shows an example of the whole structure of the image processing system by the 3rd Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the 1st image processing apparatus in the image processing system by the 3rd Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the 2nd image processing apparatus in the image processing system by the 3rd Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the switching control apparatus in the image processing system by the 3rd Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the switching control apparatus in the image processing system by the 3rd Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the 4th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the 4th Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the 5th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the 5th Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the 6th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the 6th Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the modification of the 6th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the modification of the 6th Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the 7th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the 7th Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the modification of the 7th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the modification of the 7th Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the 8th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the 8th Embodiment of this invention. It is a block diagram which shows an example of a detailed structure of the image processing apparatus in the image processing system by the modification of the 8th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the image processing apparatus in the image processing system by the modification of the 8th Embodiment of this invention. It is a block diagram which shows an example of the detailed structure of the switching control apparatus in the image processing system by the other modification of the 8th Embodiment of this invention. It is a flowchart which shows an example of the process sequence of the switching control apparatus in the image processing system by the other modification of the 8th Embodiment of this invention. In an example in which three points are detected on the xy plane of the orthogonal coordinate system, and the parameter lines corresponding to the three points intersect at different coordinates on the ρ-θ plane of the polar coordinate system in the Hough transform. It is explanatory drawing which shows a xy plane. In an example in which three points are detected on the xy plane of the orthogonal coordinate system, and the parameter lines corresponding to the three points intersect at different coordinates on the ρ-θ plane of the polar coordinate system in the Hough transform. It is explanatory drawing which shows (rho) -theta plane. An example in which three points are detected on the xy plane of the orthogonal coordinate system, and the parameter lines corresponding to the three points intersect at the same coordinates on the ρ-θ plane of the polar coordinate system in the Hough transform It is explanatory drawing which shows xy plane in. An example in which three points are detected on the xy plane of the orthogonal coordinate system, and the parameter lines corresponding to the three points intersect at the same coordinates on the ρ-θ plane of the polar coordinate system in the Hough transform It is explanatory drawing which shows (rho) -theta plane. It is explanatory drawing which showed the example from which the length of the line of a pixel unit changes with inclination. It is explanatory drawing for demonstrating the content processed with the image processing apparatus of related technology. It is explanatory drawing which shows the example in case a contour line is extracted in the image processing apparatus of related technology.

Explanation of symbols

1, 100, 200 Image processing system 10 Input device
11 Image Input Unit 12 Character Input Unit 20 Storage Device 21 Threshold Storage Unit
22 Parameter space storage unit 23 Image storage unit
30 Output Device 31 Image Output Unit
32 Display control unit 40 Image processing device 40a Candidate line drawing control means 40b Parameter space drawing means 40c Sorting means
40d Candidate line identification index calculation means 41 Binarization processing unit 42 Parameter calculation unit
43 parameter line drawing unit 44 intersection detection unit 45 overlap count measurement unit 46 leading candidate line drawing unit 47 passage determination unit (candidate line surrounding area determining means)
48 Weight calculation unit 49 Number measurement unit 51 Real line determination unit (real line determination control means)
52 Real line image generator
53 Module Control Unit 142 All Candidate Line Drawing Unit (Candidate Line Drawing Control Unit)
240 first image processing device 260 second image processing device 270 switching control device 270a first calculation amount estimation unit 270b second calculation amount estimation unit 271 binarization processing unit 272 parameter calculation calculation amount estimation unit
273 Parameter line drawing calculation amount estimation unit 274 Intersection detection calculation amount estimation unit 275 Overlap count calculation amount calculation unit 276 Possible candidate line drawing calculation amount estimation unit 277 Passage determination calculation amount estimation unit 278 Weight calculation calculation amount estimation unit 279 Number measurement calculation Quantity estimation unit 281 Real line determination calculation amount estimation unit 282 All candidate line drawing calculation amount estimation unit 283 All passage determination calculation amount estimation unit
284 Total weight calculation calculation amount estimation unit
285 Total number weighing calculation amount estimation unit 286 Real line determination calculation amount estimation unit
287 Method selection unit (calculation pattern selection means)
341 Smoothing processing unit 442 Expansion / contraction processing unit 541 Image reduction processing unit 554 Parameter fine adjustment control unit 643 All candidate line leading candidate line drawing unit 742 Parameter space coarse-graining processing unit 943 Pixel value weighted parameter line drawing unit 945 Pixel value Weighted overlap number metric unit 948, 1044 Pixel value weighted weight calculation unit 1173 Pixel value weighted parameter line drawing calculation amount estimation unit 1175 Pixel value weighted overlap number metric calculation amount estimation unit 1178 Pixel value weighted weight calculation calculation amount estimation 1184 All pixel value weighted weight calculation calculation amount estimation unit

Claims (22)

An image processing apparatus that represents a line that can pass through pixels in an image by an equation including a specific parameter, and that performs a process of extracting a line from the image using a parameter space portion having the parameter of the equation as a coordinate axis. And
Candidate line drawing control means for performing control for drawing each candidate line in the image corresponding to the coordinate value based on the coordinate value taken by the combination of the parameters in the parameter space portion;
Candidate line peripheral region determining means for determining, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass;
Candidate line identification index calculation for calculating an inclination weight indicating an inclination of a tangent line of the candidate line for each passing pixel and calculating a total value of the inclination weights for each of the candidate lines. Means,
Based on the total value of each of the candidate lines calculated by the candidate line identification index calculating means, the coordinate value of one parameter combination in the parameter space corresponding to one candidate line Search for the presence or absence of coordinate values of other parameter combinations corresponding to the other candidate lines that exceed the total value of the one candidate line in the parameter value peripheral area around In the case where the total value of the other candidate lines exceeding the total value of the one candidate line is not present, real line determination control means for performing control for determining the one candidate line as a real line;
An image processing apparatus comprising: The image processing apparatus according to claim 1.
The candidate line drawing control means includes:
A parameter space drawing means for drawing a parameter space graphic element corresponding to the pixel for each of the pixels in the parameter space;
Based on the number of intersections at the intersection where each of the parameter space graphic elements intersects, the number of intersections of other intersections around the intersection, and the preset number of intersection thresholds, Screening means for selecting potential candidates for coordinate values of the parameter combinations in the parameter space corresponding to candidate lines in the image;
Candidate line drawing means for drawing a candidate line corresponding to within the image for each coordinate value in the potential candidate area around the coordinate value of the potential candidate selected by the selection means;
An image processing apparatus comprising: The image processing apparatus according to claim 2,
Image reduction processing means for performing processing for reducing the image;
After the real line is determined by the first reduced image obtained by reducing the image at the first magnification by the image reduction processing means, the determined real line is set as a new new candidate line, and the new candidate line is determined. A new search is performed for the area around the new candidate line around the area, and a new real line is newly generated using the parameter space in the second reduced image reduced at the second magnification higher than the first magnification. A first adjustment control unit that performs determination, repeatedly performs the reduction and the determination of the new real line in a stepwise manner, and performs adjustment control of the determination of the real line;
An image processing apparatus further comprising: The image processing apparatus according to claim 3.
A discretization width adjusting means for adjusting a discretization width of each coordinate axis of the parameter space portion;
After the actual line is determined using the parameter space portion having the first discretization width in which the discretization width is increased by the discretization width adjusting unit, the determined real line is used as a new new candidate line. A new search is performed for the area around the new candidate line around the new candidate line, and a new real line is newly generated using the parameter space having a second discretization width smaller than the first discretization width. Determining, changing the discretization width and determining the new real line in a stepwise manner, second adjustment control means for performing adjustment control of the determination of the real line;
An image processing apparatus further comprising: The image processing apparatus according to claim 4.
Pixel value weight calculating means for calculating a specific pixel value weight based on the pixel value of the image before extracting the candidate area;
The candidate line identification index calculating means includes:
A pixel value weighted gradient weight is calculated for each passing pixel based on the pixel value weight and the gradient weight, and a total value of the pixel value weighted gradient weights is calculated for each candidate line. An image processing apparatus. The image processing apparatus according to claim 5.
The parameter space drawing means is
When drawing the parameter space graphic element corresponding to the pixel in the parameter space part, the pixel value weight corresponding to the pixel is the element of the element whose coordinates of the parameter space graphic element and the parameter space part match Give it to a number and draw
The selecting means includes
Based on the total number of the pixel value weights at the intersection where each of the parameter space graphic elements intersects and a preset pixel value weight total number threshold value, the intersection of the parameter space part and the surrounding area of the intersection An image processing apparatus characterized by selecting a promising candidate area in the parameter space. The image processing apparatus according to claim 6.
The pixel value weight calculation means includes:
A pixel value weight is calculated by functionally converting the pixel value;
The candidate line identification index calculating means includes:
An image processing apparatus that calculates a total value of the inclination weights with pixel value weights for each of the candidate lines. The image processing apparatus according to claim 7.
A first calculation pattern according to a first calculation pattern when performing drawing processing on the parameter space portion by the parameter space portion drawing means and drawing processing of each selected candidate line in the image by the candidate line drawing means. Of selecting the calculation pattern with the smaller calculation amount based on the calculation amount of the second calculation amount and the second calculation amount based on the second calculation pattern when only the drawing processing of each candidate line in the image is performed An image processing apparatus further comprising calculation pattern selection control means for performing the above. The image processing apparatus according to claim 8.
Candidate area extraction means for extracting candidate areas from which the line can be extracted from the image,
The candidate line drawing control means draws each candidate line in the candidate area,
The real line determination control means determines a real line from each of the candidate lines in the candidate area. The image processing apparatus according to claim 9.
Expansion and contraction processing means for performing a specific number of times one or both of the processes for expanding and contracting the candidate region,
The candidate line drawing control means includes:
An image processing apparatus, wherein the candidate line is drawn based on the image of the candidate area processed by the expansion / contraction processing means. The image processing apparatus according to claim 10.
Before extracting the candidate area, further comprising a smoothing processing means for smoothing the image;
The candidate line drawing control means includes:
An image processing apparatus, wherein the candidate line is drawn based on the image of the candidate area smoothed by the smoothing processing means. A first image processing apparatus for extracting lines from the input image by a first process;
A second image processing device for extracting lines from the image by a second process different from the first process;
A switching control device that performs switching control between the first image processing device and the second image processing device;
An output device that outputs an image processing result of one or both of the first and second image processing devices that are switch-controlled by the switching control device;
Including
The second image processing apparatus includes:
A line that passes through a pixel in the image is represented by an equation including a specific parameter, and the coordinate corresponding to the coordinate value is based on a coordinate value taken by a combination of the parameters in the parameter space portion with the parameter of the equation as a coordinate axis. Candidate line drawing control means for performing control for drawing each candidate line in the image;
Candidate line peripheral region determining means for determining, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass;
Candidate line identification index calculation for calculating an inclination weight indicating an inclination of a tangent line of the candidate line for each passing pixel and calculating a total value of the inclination weights for each of the candidate lines. Means,
Based on the total value of each of the candidate lines calculated by the candidate line identification index calculating means, the coordinate value of one parameter combination in the parameter space corresponding to one candidate line Search for the presence or absence of coordinate values of other parameter combinations corresponding to the other candidate lines that exceed the total value of the one candidate line in the parameter value peripheral area around In the case where the total value of the other candidate lines exceeding the total value of the one candidate line is not present, real line determination control means for performing control for determining the one candidate line as a real line;
An image processing system comprising: The image processing system according to claim 12.
The first image processing apparatus includes:
A line passing through a pixel in the image is represented by an equation including a specific parameter, and a parameter space graphic element corresponding to the pixel is drawn for each of the pixels in a parameter space portion having the parameter of the equation as a coordinate axis. Parameter space drawing means;
Based on the number of intersections at the intersection where each of the parameter space graphic elements intersects, the number of intersections of other intersections around the intersection, and the preset number of intersection thresholds, Screening means for selecting potential candidates for coordinate values of the parameter combinations in the parameter space corresponding to candidate lines in the image;
Candidate line drawing means for drawing a candidate line corresponding to within the image for each coordinate value in the potential candidate area around the coordinate value of the potential candidate selected by the selection means;
Candidate line peripheral region determining means for determining, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass;
Candidate line identification index calculation for calculating an inclination weight indicating an inclination of a tangent line of the candidate line for each passing pixel and calculating a total value of the inclination weights for each of the candidate lines. Means,
Based on the total value of each of the candidate lines calculated by the candidate line identification index calculating means, the coordinate value of one parameter combination in the parameter space corresponding to one candidate line Search for the presence or absence of coordinate values of other parameter combinations corresponding to the other candidate lines that exceed the total value of the one candidate line in the parameter value peripheral area around In the case where the total value of the other candidate lines exceeding the total value of the one candidate line is not present, real line determination control means for performing control for determining the one candidate line as a real line;
An image processing system comprising: The image processing system according to claim 13.
The switching control device includes:
In the case where the drawing process in the parameter space part by the parameter space part drawing unit of the first image processing apparatus and the drawing process of each selected candidate line in the image by the candidate line drawing unit are performed. First calculation amount estimation means for estimating a first calculation amount according to the first calculation pattern;
Second calculation amount estimation means for estimating a second calculation amount based on a second calculation pattern when only drawing processing of each candidate line in the image is performed;
The first calculation amount in the first calculation amount estimation means and the second calculation amount in the second calculation amount estimation means are compared, and the calculation pattern with the smaller calculation amount is selected. A calculation pattern selection means;
An image processing system comprising: An image processing method in which a line that passes through pixels in an image is represented by an equation including a specific parameter, and a computer extracts a line from the image by using a parameter space section having the parameter of the equation as a coordinate axis Because
A candidate line drawing control step for performing control for drawing each candidate line in the image corresponding to the coordinate value, based on the coordinate value taken by the combination of the parameters in the parameter space unit,
A candidate line peripheral region determining step in which the computer determines, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass;
A candidate for calculating an inclination weight indicating an inclination of a tangent line of the candidate line for each passing pixel, and calculating a total value of the inclination weights for each of the candidate lines, wherein the computer is an index for identifying the candidate line. A line identification index calculating step;
Based on the total value of each of the candidate lines calculated in the candidate line identification index calculating step, the computer sets one parameter in the parameter space portion corresponding to one candidate line. The parameter value surrounding the coordinate value of the combination is searched for the presence or absence of the coordinate value of the other parameter combination corresponding to the other candidate line that exceeds the total value of the one candidate line, and the parameter Real line determination control for performing control to determine one candidate line as a real line when there is no total value of the other candidate lines that exceeds the total value of one candidate line in a value peripheral region Steps,
An image processing method comprising: The image processing method according to claim 15, wherein
In the candidate line drawing control step,
A parameter space drawing step for drawing a parameter space graphic element corresponding to the pixel for each of the pixels in the parameter space;
Based on the number of intersections at the intersection where each of the parameter space graphic elements intersects, the number of intersections of other intersections around the intersection, and the preset number of intersection thresholds, A selection step of selecting potential candidates for coordinate values of the parameter combinations of the parameter space portion corresponding to candidate lines in the image;
A candidate line drawing step for drawing a candidate line corresponding to the inside of the image for each coordinate value in the potential candidate area around the coordinate value of the potential candidate selected in the selection step;
And an image processing method. The image processing method according to claim 16.
An image reduction processing step in which the computer performs a process of reducing the image;
After the computer determines the real line in the first reduced image obtained by reducing the image at the first magnification in the image reduction processing step, the determined real line is set as a new new candidate line, A new search is performed for the area around the new candidate line around the new candidate line, and a new reduced image reduced at a second magnification higher than the first magnification is newly used using the parameter space. A first adjustment control step of determining a new real line, repeatedly performing the reduction and the determination of the new real line in a stepwise manner, and performing adjustment control of the determination of the real line;
An image processing method characterized by further comprising: The image processing method according to claim 17,
The computer adjusts the discretization width of each coordinate axis of the parameter space portion to adjust the discretization width;
After the computer determines the real line using the parameter space portion of the first discretization width that has been increased in the discretization width adjustment step, the determined real line is newly determined. A new candidate line, a new search is performed for a region around the new candidate line around the new candidate line, and a new one is newly created using the parameter space having a second discretization width smaller than the first discretization width. A second adjustment control step for determining a new real line, repeatedly performing the change of the discretization width and the determination of the new real line in a stepwise manner, and performing adjustment control of the determination of the real line;
An image processing method characterized by further comprising: A program that allows a computer to execute a process of extracting a line from the image by using a parameter space portion that uses the parameter of the equation as a coordinate axis, and represents a line that can pass through a pixel in the image as a specific parameter. Because
A candidate line drawing control function for performing control for drawing each candidate line in the image corresponding to the coordinate value based on the coordinate value taken by the combination of the parameters in the parameter space part;
A candidate line peripheral region determination function for determining, for each of the candidate lines, a passing pixel through which the candidate line and a candidate line peripheral region around the candidate line pass;
Candidate line identification index calculation for calculating an inclination weight indicating an inclination of a tangent line of the candidate line for each passing pixel and calculating a total value of the inclination weights for each of the candidate lines. Function and
Based on the total value of each of the candidate lines calculated by the candidate line identification index calculation function, the coordinate value of one parameter combination in the parameter space portion corresponding to one candidate line Search for the presence or absence of coordinate values of other parameter combinations corresponding to the other candidate lines that exceed the total value of the one candidate line in the parameter value peripheral area around In the case where there is no total value of the other candidate lines exceeding the total value of the one candidate line, a real line determination control function for performing control to determine one candidate line as a real line;
A program for causing a computer to execute a function including: The program according to claim 19, wherein
The candidate line drawing control function is:
A parameter space drawing function for drawing a parameter space graphic element corresponding to the pixel for each of the pixels in the parameter space;
Based on the number of intersections at the intersection where each of the parameter space graphic elements intersects, the number of intersections of other intersections around the intersection, and the preset number of intersection thresholds, A selection function for selecting potential candidates for coordinate values of the parameter combinations in the parameter space corresponding to candidate lines in the image;
A candidate line drawing function that draws a candidate line corresponding to within the image for each coordinate value in a potential candidate area around the coordinate value of the potential candidate selected by the selection function;
A program for causing a computer to execute a function including: The program according to claim 20,
An image reduction processing function for performing processing for reducing the image;
After the real line is determined by the first reduced image obtained by reducing the image at the first magnification by the image reduction processing function, the determined real line is set as a new new candidate line. A new search is performed for the area around the new candidate line around the area, and a new real line is newly generated using the parameter space in the second reduced image reduced at the second magnification higher than the first magnification. A first adjustment control function for determining, performing the reduction and the determination of the new real line in a stepwise manner, and performing adjustment control of the determination of the real line;
Is further executed by a computer. The program according to claim 21, wherein
A discretization width adjustment function for adjusting the discretization width of each coordinate axis of the parameter space portion;
After the actual line is determined using the parameter space portion of the first discretized width obtained by increasing the discretized width by the discretized width adjusting function, the determined real line is used as a new new candidate line. A new search is performed for the area around the new candidate line around the new candidate line, and a new real line is newly generated using the parameter space having a second discretization width smaller than the first discretization width. Determining, changing the discretization width and determining the new real line in a stepwise manner, and a second adjustment control function for performing adjustment control of the determination of the real line;
Is further executed by a computer.

Images