JPH08329252A - Method and device for detecting edge - Google Patents

Abstract

PURPOSE: To improve the precision of edge detection by calculating the edge intensity to detect the picture element group, where the picture element value is steeply changed in comparison with peripheral picture element values, as an edge. CONSTITUTION: A gradient calculation part 11 to which an input picture Fin is supplied, a spline curve generation part 12 to which an outline P of contours is supplied, a parameter coordicate generation part 13 which generates a parameter (t) to the spline,curve generation part 12, and an edge intensity calculation part 14 which calculates an edge intensity E in accordance with the output from the gradient calculation part 11 and that from the spline curve generation part 12 are provided. The edge direction of each position in input picture data consisting of variable density data is estimated. The gradient in each position is obtained. The edge intensity calculation part 14 takes information of the estimated edge direction as auxiliary information and calculates the edge intensity E in accordance with this auxiliary information and the gradient, and the picture element group where the picture element value is steeply changed in comparison with peripheral picture element values is detected as an edge.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an edge detection method and an edge detection device which play a basic role in image processing, and special effect processing in FA (Factory Auto) in video production such as television and movies.
The present invention relates to an edge detection method and an edge detection apparatus which are suitable for application to component recognition from a camera image in the operation (.

[0002]

2. Description of the Related Art The operation of cutting out an arbitrary partial image from an image is a basic operation of editing and synthesizing video images and generating texture and structure data in computer graphics and the like. This cutout operation requires processing such as edge detection, area extraction, and association with a known object.
In particular, when the background or the target object in the image becomes complicated, it is necessary to accurately find and track the edges forming the contour of the target object.

Here, the edge detection is a process for finding a portion in which a pixel value changes abruptly in a grayscale image.
Usually, a sharp change occurs in the contour of the target object, so the contour of the target object can be extracted from the image based on the result of edge detection. Therefore, edge detection is used in many fields as the most basic process for obtaining information about an object existing therein from an image.
In particular, the extraction of the contour of the target object is the most basic process for determining the configuration, position, etc. of the target object existing in the image by a computer or the like, and is the most important application of edge detection.

There are many publicly known methods for edge detection as described above, and in any of these methods, a local change in pixel value is examined and a portion with a large change is detected as an edge. . As a typical edge detection method, there is a method using a differential filter as a spatial filter. The edge detection method using this differential filter is a method of detecting an edge based on the principle that the differential coefficient also has a large value in a portion where the change in pixel value is large. For example, a lateral (X) direction Sobel filter as shown in FIG.
The calculation corresponding to the first-order differential in the horizontal (X) direction is performed, and when this Sobel filter is used, the output of the Sobel filter is positive or negative when an edge in the vertical (Y) direction exists. Is large, the edge in the vertical (Y) direction can be detected.

In addition to the Sobel filter as described above,
Various types of spatial filters that perform calculations corresponding to first-order or second-order differentiation are used for edge detection. These edge detection methods are described in, for example, “Fundamentals of Fundamentals” by Jain, which is a typical textbook.
Digital image processing (Fundame
ntals of Digital Image Pr
processing "and Pratt's" Digital Image Processing (Digital).
Image Processing).

Further, various methods for detecting the contour of a target object using the above-described edge detecting method have been tried. For example, a binary image in which the degree of change in pixel value, that is, the portion having a large edge strength is “1” and the other portion is “0” is generated from the edge detection result, and the binary image is thinned. A region extraction method for obtaining the contour of a target object by doing so is disclosed in Japanese Patent Application No. Hei 5-233810. Here, thinning means that pixels with a value of "1" are sequentially removed from the generated binary image, and pixels with a value of "1" are normally removed in order until the width becomes 1 pixel. Is a known procedure of extracting

On the other hand, there is also a method of detecting edges by a statistical method without using a spatial filter. For example, an edge detecting method and apparatus for calculating a variance of pixel values in the vicinity of a pixel of interest, that is, a variance of hues and detecting a place where the value is large as a boundary of a region is disclosed in Japanese Patent Application No. 5-18196.
No. 9 ”.

In addition, the document "Color Edge Detection, Eusing Vector Ordered Statistics (Trahanias, PE et al." Co.
lor edge detection using
vector order statistics ”, I
EEE Transactions on image
processing, Vo1.2, No. 2, p
p. 259-264, 1993) "describes a method for detecting edges based on a scale indicating how far a pixel value is from each other pixel value.

[0009]

However, the conventional edge detecting method as described above has a problem that all the portions where the pixel value changes abruptly are detected. For example, as shown in FIG. 11A, in the target object 110 of the image 100, when there is a color difference inside the target object 110 or an intersection of the background and the target object 110, the conventional edge detection method is used. As shown in (b), the boundary 112 inside the object was also detected, and it was not possible to obtain only the edge 111 to be originally obtained as shown in (c) of the figure.

Normally, a sharp change in pixel value is unavoidable inside the object due to various reasons such as the color difference, the pattern, and the shadow as shown in FIG. 11A. Also, there are usually various patterns and colors in the background. On the other hand, in the conventional edge detection method, a sharp change in the pixel value is detected uniformly, so that the boundary inside the object or the background pattern as described above may be erroneously detected as an edge. was there.

Therefore, in order to solve the above-mentioned problem, a weighting correction process is performed on the lightness based on the color information of the image, that is, the lightness, the saturation, and the hue, and then the weighting correction process is performed. An automatic cutting system for detecting an edge based on brightness is disclosed in Japanese Patent Application No. 1-173177. This automatic clipping system makes the contours more prominent by weighting correction processing for emphasizing changes in pixel values near the contours, for example, changes in brightness, so that erroneous patterns and color changes in foreground objects and backgrounds are detected. Edge detection is performed with reduced detection.

However, the above automatic cutting system does not consider performing the optimum weighting correction processing for each individual portion on the contour. Further, it is rather rare that the pattern, color, etc. are uniform along the contour of the object.
For this reason, even the above automatic cutting system cannot prevent the mixing of edges other than the originally desired edge.

In addition, an edge extraction device that detects edges by converting the pixel values so as to increase the contrast at the boundary between the foreground and the background by using the histogram of the image is disclosed in "Patent application". Flat 1-76
170 ". However, even with the above-mentioned edge extraction device, although it is effective in detecting sensitive edges, it is not possible to prevent mixing of edges other than the originally desired edge.

On the other hand, by limiting the place where the edge detection is performed on the image, the edge detection in the unnecessary place can be omitted, and the area extracting device which minimizes the mixing of the edge in the place other than the contour is disclosed in "Patent application". No. 3-176780 ". In this area extraction device, the outline of the target object is manually input, and the edge detection is performed only within the outline of the target object, thereby preventing the generation of extra edges and performing the edge detection. Further, a softkey generation device configured to manually input the outline of a target object and perform edge detection only near the boundary based on the outline information is disclosed in Japanese Patent Application No. 5-236347. ing.

However, in the area extracting device and the soft key generating device, only the area in which an erroneous edge is detected is narrowed, and it is impossible to prevent the mixing of edges other than the contour. For example, as shown in FIG. 12A, in the image 200, when edge detection is performed using the outline 220 of the contour of the target object 210 input by the operator, even if the range in which the edge detection is performed is limited, the limitation is performed. Has a limit, and as shown in FIG. 12B, unnecessary edges 221a, 221b, 221c have been erroneously detected.

Here, by increasing the precision of the outline 220, unnecessary edges 221a, 221b, 221 are obtained.
Although it is possible to gradually reduce the number and length of c, it is necessary to manually give the outline 220 of the contour with high precision, which is a very difficult task. To solve this problem, when the operator is slowly moving the input means such as a tablet or mouse, the outline is controlled so that the outline thickness becomes thin. An image editing apparatus that detects an edge by limiting the range in which edge detection is performed by controlling the outline to be thick is disclosed in "Japanese Patent Application No. Hei 1-180674". However, it was not possible to prevent mixing of unnecessary edges.

In order to solve the above-mentioned problem, for example, the edge direction is estimated from a rough outline manually input, and only the edges that match the estimated edge direction are selectively detected, whereby the above-mentioned problem is solved. There is an edge detection method for detecting an edge by largely removing such an unnecessary edge.

In the above-mentioned edge detection method, FIG.
~ Robinson as shown in (d)
By using a filter group called “3-level”, only edges in a specific direction are detected. The filter group shown in (a) to (d) of FIG. 13 is an extension of the Sobel filter shown in FIG. 10 and detects edges in directions other than horizontal and vertical directions. For example, the filter shown in (a) of FIG. 13 outputs a large value when the pixel value on the right side of the image is large and the pixel value on the left side of the image is small. That is, the filter shown in (a) of FIG. 13 detects an edge in the + 0 ° direction. Further, the filters shown in (b), (c), and (d) of FIG.
Edges in the directions of 45 °, 90 °, and 135 ° are detected.

Further, in the above edge detecting method, the above-mentioned FIG.
In addition to the filter groups shown in (a) to (d) of 3 above, in a direction other than the direction of the edge detected by the filter group,
A filter group that detects edges in 45 ° increments is used. Therefore, in the above edge detection method, a filter for detecting edges in all eight directions is used.

Here, in addition to the Robinson three-level filter shown in FIGS. 13A to 13D, as a filter for detecting an edge in a specific direction, the above-mentioned "Digital" by Pratt. Image processing (Digital Image Process)
ng) ”described in Prewit
There are a compass filter of t), a Kirsch filter, and the like.

However, in the above-mentioned filter for detecting edges in a specific direction (hereinafter, referred to as direction-selective filter) such as the Robinson three-level filter, the Prewitt compass filter, and the Kirsch filter, the direction is determined. There is a drawback that the selectivity, that is, the directivity is weak.

More specifically, for example, when the original strength of the edge is K and the angle between the edge direction to be detected and the actual edge direction is θ, the output of the direction-selective filter, that is, the edge strength. E is given by the relational expression of E = K cos θ.

In this relational expression, the edge strength E of the unnecessary edge does not become "0" unless the direction of the unnecessary edge differs from the direction of the edge to be originally detected by 90 °, that is, the mixing of the unnecessary edge is completed. It indicates that it cannot be removed. Therefore, as shown in FIG. 14, the directivity B required for practical use needs to be stronger than the directivity A in the direction-selective filter. Since it was not possible to provide a selective filter, it was not possible to prevent the mixing of unnecessary edges.

Further, in the direction-selective filter, when it is desired to make the edge detection direction finer, for example, when the edge detection direction is made finer in 15 ° units than in the 45 ° units as described above. Had a large filter. That is, the positive and negative portions of the filter coefficient must be arranged at right angles to the edge detection direction, but the edge detection direction is 15 °.
If it is specified in detail like a unit, a small filter of about 3 × 3 becomes insufficient to represent the angle of the detection direction, and the filter becomes large. As a result, the ability to detect fine edges deteriorates, and inconvenience occurs in that edges cannot be dealt with in all directions unless a large number of filters are used.

Therefore, the present invention has been made in view of the above-mentioned conventional circumstances, and has the following objects. That is, it is an object of the present invention to provide an edge detection method and an edge detection device which prevent unwanted mixing of edges and improve the edge detection accuracy.

Another object of the present invention is to provide an edge detecting method and an edge detecting apparatus having strong directivity in the edge detecting direction. It is another object of the present invention to provide an edge detection method and an edge detection device that can easily obtain a good edge image.

Another object of the present invention is to provide an edge detecting method and an edge detecting apparatus for obtaining a good edge image at high speed.

[0028]

In order to solve the above-mentioned problems, the edge detection method according to the present invention is such that the pixel value in the input image data composed of grayscale data changes sharply as compared with the surroundings. An edge detection method for detecting a group of pixels as an edge, estimating the edge direction of each position in the input image data, obtaining the gradient at each position, and using the information of the estimated edge direction as auxiliary information, It is characterized in that edge strength is calculated from auxiliary information and the gradient to detect an edge.

The edge detecting method according to the present invention is characterized in that the edge strength is calculated by obtaining the cosine of the angle formed by the estimated edge direction and the gradient. Further, the edge detection method according to the present invention, an angle θ formed by the estimated edge direction and the gradient,
It is characterized in that the edge strength E is calculated by a calculation of E = C | G | cos ^S (θ) with a proportionality constant C, the gradient size G, and an index S that determines directivity.

In the edge detecting method according to the present invention, the estimated edge direction which is a vector obtained by normalizing the cosine of the angle formed by the estimated edge direction and the gradient and the length thereof are set to "1". It is characterized in that it is obtained as an inner product with the normalized gradient.

Further, the edge detecting method according to the present invention is characterized in that the edge strength is calculated from the auxiliary information and the gradient using an edge strength information table calculated in advance. An edge detection device according to the present invention is an edge detection device that detects, as an edge, a pixel group in which pixel values change sharply compared to the surroundings from input image data composed of grayscale data. Estimating the edge direction of each position in the image data, estimating means using the information of the estimated direction as auxiliary information, gradient calculating means for obtaining the gradient of each position in the input image data, and the auxiliary obtained by the estimating means. It is characterized by comprising edge strength calculation means for calculating edge strength from information and the gradient obtained by the gradient calculation means, and detection means for detecting an edge from the edge strength obtained by the edge strength calculation means.

Further, the edge detecting apparatus according to the present invention is characterized in that the edge strength calculating means calculates the edge strength by obtaining a cosine of an angle formed by the estimated edge direction and the gradient. Further, in the edge detection device according to the present invention, the edge strength calculation means is configured to form an angle θ between the estimated edge direction and the gradient.
, The proportional constant C, and the magnitude G of the gradient,
It is characterized in that the edge strength E is calculated by a calculation of E = C | G | cos ^S (θ) with an index S that determines the directivity.

Further, in the edge detecting apparatus according to the present invention, the edge strength calculating means has the estimated edge direction which is a vector obtained by previously normalizing a cosine of an angle formed by the estimated edge direction and the gradient, It is characterized in that it is obtained as an inner product with the above-mentioned gradient in which the length is normalized to "1".

Further, the edge detecting apparatus according to the present invention comprises an information table of edge strength calculated in advance, and the edge strength calculating means calculates the edge strength from the auxiliary information and the gradient using the information table. It is characterized by

[0035]

In the edge detecting method according to the present invention, the edge direction of each position in the input image data composed of grayscale data is estimated. Also, the gradient at each position is obtained. Then, the estimated edge direction information is used as auxiliary information, edge strength is calculated from the auxiliary information and the gradient, and a pixel group in which the pixel value changes sharply compared with the surroundings is detected as an edge.

Further, in the edge detecting method according to the present invention,
The edge strength is calculated by obtaining the cosine of the angle formed by the estimated edge direction and the gradient. Further, in the edge detection method according to the present invention, the angle θ formed by the estimated edge direction and the gradient, and the proportional constant C
Then, with the magnitude G of the gradient and the index S that determines the directivity, the edge strength E is calculated by a calculation of E = C | G | cos ^S (θ).

Further, in the edge detecting method according to the present invention,
The cosine of the angle formed by the estimated edge direction and the gradient is obtained as the inner product of the estimated edge direction, which is a previously normalized vector, and the gradient whose length is normalized to "1".

Further, in the edge detecting method according to the present invention,
The edge strength is calculated from the auxiliary information and the gradient using the edge strength information table calculated in advance. In the edge detecting device according to the present invention, the estimating means estimates the edge direction of each position in the input image data composed of the grayscale data, and uses the information of the estimated direction as auxiliary information. The gradient calculating means calculates the gradient at each position in the input image data. The edge strength calculating means calculates the edge strength from the auxiliary information obtained by the estimating means and the gradient obtained by the gradient calculating means. The detection unit detects, as an edge, a pixel group in which the pixel value changes sharply compared with the surroundings, from the edge strength obtained by the edge strength calculation unit.

Further, in the edge detecting device according to the present invention,
The edge strength calculating means calculates the edge strength by obtaining the cosine of the angle formed by the estimated edge direction and the gradient. Further, in the edge detection device according to the present invention, the edge strength calculation means includes an angle θ formed by the estimated edge direction and the gradient, and a proportional constant C.
Then, with the magnitude G of the gradient and the index S that determines the directivity, the edge strength E is calculated by a calculation of E = C | G | cos ^S (θ).

In the edge detecting device according to the present invention,
The edge strength calculation means calculates the cosine of the angle formed by the estimated edge direction and the gradient from the estimated edge direction, which is a vector that is normalized in advance, and the gradient whose length is normalized to “1”. Calculate as the dot product.

Further, in the edge detecting device according to the present invention,
The edge strength calculation means calculates the edge strength from the auxiliary information and the gradient using an edge strength information table calculated in advance.

[0042]

An embodiment of the present invention will be described below with reference to the drawings. The edge detection method according to the present invention is carried out by an edge detection device 10 as shown in FIG. 1, for example. The edge detection device 10 is an application of the edge detection device according to the present invention.

That is, as shown in FIG. 1, the edge detecting apparatus 10 includes a gradient calculating section 11 to which an input image F _in is supplied, a spline curve generating section 12 to which a contour outline P is supplied, and an auxiliary variable. A parameter coordinate generation unit 13 that generates t for the spline curve generation unit 12, and an edge strength calculation unit 14 that calculates an edge strength E from the output from the gradient calculation unit 11 and the output from the spline curve generation unit 12 are provided. ing. The output from the spline curve generator 12 is also supplied to the gradient calculator 11.

In addition, the gradient calculation unit 11 is shown in FIG.
As shown in FIG. 4, the filter calculator 111 and the filter calculator 112 are provided to which the input image F _in is supplied. The filter calculator 111 uses the Sobel filter S _{X in the} X direction.
Is applied to the input image F _in , and the filter calculator 112 applies the Y-direction Sobel filter S _Y to the input image F _in . The outputs from the filter calculator 111 and the filter calculator 112 are supplied to the edge strength calculator 14.

Further, the edge strength calculation section 14 is supplied with a preset direction selection index s, and the edge strength calculation section 14, as shown in FIG.
The gradient table T _G referenced using the output from the gradient calculation unit 11, the coordinate rotation unit 412 to which the output from the spline curve generation unit 12 is supplied, the result obtained by the gradient table T _G , and the coordinate rotation unit 4
The inner product calculation unit 141 to which the output from 12 is supplied, the cosine table T _C referred to by using the supplied direction selection index s and the output from the internal calculation unit 141, and the result obtained by the cosine table T _C. A multiplier 143 for calculating the edge strength E from the result obtained by the gradient table T _G
It has and.

Here, FIG. 4 is a flowchart showing the edge detection processing in the edge detection device 10. Hereinafter, a specific description will be given with reference to FIGS. 1 to 3 and 4. An input image F _in composed of grayscale data is input to the edge detection device 10. On the other hand, the edge detection device 10
For example, the outline P of the contour is input by the operator using the tablet while observing the input image F _in displayed on the display (not shown).

The gradient calculator 11 calculates the input image F
_In is taken in (step S1). The detailed description of the gradient calculator 11 will be described later. On the other hand, the spline curve generation unit 12 calculates the outline P of the contour input by the operator as described above into the coordinate sequence (x0, y0), (x
1, y1), (x2, y2), ... (Step S2). Then, the spline curve generator 1
2 is the coordinate sequence (x0, y0), (x1,
A smooth curve P (t) passing through y1), (x2, y2), ... Is generated (step S3).

Specifically, first, the curve P (t)
Is expressed in a cubic spline format. That is, the curve P (t) is the auxiliary variable t and the cubic polynomial x of the auxiliary variable t.
It has (t) and y (t), and is expressed by an equation P (t) = (x (t), y (t)). Then, the auxiliary variable t is "0"-
When it changes to “1”, the locus of the curve P (t),
That is, the spline curve P (t) has a coordinate sequence (x0, y
0), (x1, y1), (x2, y2), ... Are smoothly connected to make a round of the contour. Such spline curves are used in a wide range of fields including CAD.

Therefore, the spline curve generation unit 12 determines that the spline curve P (t) is the coordinate sequence (x0, y0), (x
1, y1), (x2, y2), ... Coordinate sequences (x0, y0), (x1) in which coefficients of cubic polynomials x (t) and y (t) are sequentially taken in so as to smoothly connect them. , Y1),
(X2, y2), ... (hereinafter, coordinate sequence (xn, yn)
Say ) Based on. The method of determining this coefficient is
For example, a representative document, Farin.
"Curves and Surfaces for Computer Aided Geometric Design (Curves and surfaces for for
computer aided geometric
The method of determining the coefficient described in "Design" is applied. The method of determining this coefficient is that, given a point cloud,
This is a publicly known method for determining a coefficient that passes through the point group.

For example, when the input image F _in is the image 250 as shown _in FIG. 5A, that is, when the circular object 260 exists near the center of the image 250, the spline curve generating section 12 determines that the circular shape is circular. The point group 271 _xy on the outline specified by the operator along the periphery of the object 260 is captured as a coordinate sequence (xn, yn). Then, the spline curve generation unit 12 uses the curve 27 that interpolates the point group 271 _xy.
A coefficient such that 0 passes through the point group 271 _xy is determined to generate the spline curve P (t).

As described above, the spline curve generator 1
The spline curve P (t) generated by 2 is supplied to the gradient calculation unit 11. In addition, the spline curve generation unit 12 uses the generated spline curve P (t) as an outline of the contour and performs the following process for each pixel on the spline curve P (t) (step S4). At this time, the parameter coordinate generation unit 13 changes the auxiliary variable t of the curve P (t) from "0" to "1" with a small step size and generates it for the spline curve generation unit 12.

The process of step S4 described below is as follows.
This is a loop process in which each pixel on the spline curve P (t) is sequentially traced by the auxiliary variable t that is gradually changed from the parameter coordinate generation unit 13, and is repeated for each pixel (x, y). First, the spline curve generation unit 12 sets the auxiliary variable t from the parameter coordinate generation unit 13 to “P
(T) = (x (t), y (t)) ”to obtain the X and Y coordinates of the target pixel (hereinafter referred to as the target pixel (x, y)). Then, the spline curve generation unit 12 supplies the obtained target pixel (x, y) to the gradient calculation unit 11.

Further, as shown in FIG. 5 (b), the spline curve generating section 12 finds the tangential direction of the spline curve P (t) at the obtained target pixel (x, y). This tangential direction is used as the approximate direction of the contour at the target pixel (x, y). That is, the tangential direction is the velocity vector V obtained by differentiating the spline curve P (t).
Since it is equal to the direction of (t), this velocity vector V (t)
Is taken as the tangential direction, and V (t) = d / dtP (t) is obtained. Then, the obtained velocity vector V (t)
Is normalized to "1", and the tangent vector T is calculated by the formula T = V (t) / | V (t) | (step S4.1), and the calculated tangent vector T is calculated by the edge strength calculation unit. Supply to 14.

Next, the tangent vector T obtained by the spline curve generator 12 as described above is used as auxiliary information, and the edge detection processing described below is performed using this auxiliary information (step S4.2). First, the edge strength calculator 14
Calculates a normal vector N obtained by rotating the tangent vector T from the spline curve generating unit 12 by 90 ° (step S
4.2.1). This rotation has a positive counterclockwise direction with respect to the traveling direction of the locus of the spline curve P (t).

That is, in the edge strength calculation unit 14, as shown in FIG. 3, the coordinate rotation unit 142 converts the tangent vector T from the spline curve generation unit 12 into a normal vector N rotated counterclockwise. . The processing in the coordinate rotation unit 142 is very simple, only by multiplying the coefficients by 1 and -1.

When the processing for obtaining the normal vector N as described above is expressed by using an arithmetic expression, the normal vector N is obtained by using the matrix R shown in Formula 1.

[0057]

[Equation 1]

N = RT. The normal vector N thus obtained is
Since it is the normal vector of the spline curve P (t), if the direction of the edge and the estimated contour direction are parallel,
The gradient and the normal vector N of the spline curve P (t), which will be described later, perpendicular to each of them are also parallel to each other.

Next, the gradient calculator 11 described above
Is the target pixel (x, from the spline curve generation unit 12
y), the pixel value I (x, of the input image F _in captured in the process of step S1 for the region of 3 × 3 pixels
y) is read, and the gradient (hereinafter, referred to as a gradient) G of the image at the target pixel (x, y) is obtained.

That is, in the gradient calculator 11, as shown in FIG.
Applies the Sobel filter S _X in the X direction to the read pixel value I (x, y) of the input image F _in . Further, the filter calculator 112 applies the Sobel filter S _Y in the Y direction to the read pixel value I (x, y) of the input image F _in . Then, the gradient calculator 11 outputs the outputs Dx and D of the filter calculator 111 and the filter calculator 112.
The y is combined and supplied to the edge strength calculation unit 14 as the image gradient G (= (Dx, Dy)) at the target pixel (x, y) (step S 4.2.2).

The gradient G obtained as described above is a vector indicating the direction of maximum inclination when the light and shade of the image is taken as the altitude, and is perpendicular to the edge direction. That is, the gradient G can be expressed by the following equation for the pixel value I (x, y) of the input image F _in : G = gradI (x, y) = (Dx, Dy)

Next, the edge strength calculation unit 14 calculates the length l of the gradient G obtained by the gradient calculation unit 11.
EnG and normG normalized so that the length lenG of the gradient G becomes "1" are obtained (step S4.2.3.). Here, the edge strength calculation unit 14 includes a gradient table T _G, as shown in FIG. In this gradient table T _G , the length lenG of the gradient G and the normalized no
The rmG and the result calculated in advance using the arithmetic expression shown in [Equation 2] are stored.

[0063]

[Equation 2]

More specifically, the gradient table T _G receives the gradient G (= (Dx, Dy)) obtained by the gradient calculator 11 as an input, and normalizes the length lenG of the gradient G. nor
It is a table which outputs mG. In addition, the accuracy of the pixel value is usually 8 bits, and the same accuracy is used for the output Dx in the X direction and the output Dy in the Y direction. Therefore, the gradient table T _G includes the output Dx in the X direction and the output Dy in the Y direction. A total of 16 bits of the output Dy are input. Therefore, one element of the gradient table T _G is assigned to an arbitrary combination of the output Dx in the X direction and the output Dy in the Y direction. Therefore, for each element of the gradient table T _G , the value of the length lenG of the gradient G and the normG obtained by normalizing the value
And the value of are stored. By using such a gradient table T _G , the result can be easily obtained without performing a complicated calculation such as a square root. Further, since the input to the gradient table T _G is 16 bits, the table size, that is, the number of elements is “65536”, which can be sufficiently applied to the current computer.

The gradient table T _G as described above
Thus, the length lenG of the gradient G (= (Dx, Dy)) of the gradient G (= (Dx, Dy)) obtained by the gradient calculation unit 11 and normG obtained by normalizing the length lenG are obtained. Then, the length len of the obtained gradient G
The normG obtained by normalizing G is supplied to the inner product calculating unit 141, and the length lenG of the gradient G is supplied to the multiplier 143.

Finally, the edge strength calculator 14 calculates the edge strength E (x, y) of the contour at the target pixel (x, y) from the normal vector N and the gradient table T _G obtained as described above. Length of gradient G len
G, normG that normalizes it, and direction selection index s
Then, E (x, y) = lenG (normG · N) ^{S is} obtained.

That is, in the edge strength calculation unit 14, as shown in FIG.
Is the normal vector N and the length len of the gradient G
Inner product “normG · N” with normG normalized to G
Is obtained by multiplication. The result is the cosine “cos θ” of the angle θ formed by the gradient G and the normal vector N.

Next, using the cosine table T _C , cos ^S θ is obtained from the supplied direction selection index s. This cosine table T _C is the inner product calculation unit 1
41 is a table in which the cosine “cos θ” obtained in step 41 is input, and the S power of the cosine “cos θ” is output.
The cosine “cos” output from the cosine table T _C
The Sth power of (θ) (= cos ^S θ) is supplied to the multiplier 143. Here, in this embodiment, since the direction selection index s is a preset fixed value, the input to the cosine table T _C is the result of the inner product obtained by the inner product calculation unit 141, and the cosine table T _C is a small table whose input is about 8 bits. By using such a cosine table T _C , it is possible to easily obtain the S-th power “cos ^S θ” of the cosine without performing complicated calculations.

Finally, the multiplier 143 determines the length l of the gradient G from the gradient table T _G.
EnG is multiplied by the cosine Sth power “cos ^S θ” from the cosine table T _C , and the multiplication result is output as the edge strength E. Here, the direction selection index s supplied to the edge strength calculation unit 14 is an index that determines how selectively and selectively reacts only to edges in a specific direction. The larger the value of the direction selection index s, the stronger the directivity. Therefore, it is possible to more strongly prevent the mixing of unnecessary edges, but the desired direction, that is, the estimated edge direction and the actual edge direction are If it deviates even a little, the edge strength is attenuated. Therefore, the direction selection index s is usually set to a value between 2 and 8.

For example, as shown in FIG. 6, when the direction selection index s increases, the gradient G and the normal vector N
In the same direction, that is, when the value of the inner product “normG · N” is close to “1”, the edge strength E (x, y) becomes a large value. That is, high directivity can be obtained. Also,
When the direction selection index s is "1", the same directivity as that of the Robinson three-level filter, the Prewitt compass filter, and the Kirsch filter described above is obtained.

Each pixel (x, x on the spline curve P (t) giving the outline of the processing of step S4 as described above
y). Therefore, the edge strength calculator 14
Then, as shown in FIG. 7, the edge intensity image F _out obtained from the input image F _in is output (step S5).
As described above, in the present embodiment, in the process of obtaining the edge strength E, the cosine “cos θ” is directly obtained by calculating the inner product without directly obtaining the angle formed by the gradient G and the normal vector N. There is no need to perform complicated processing such as calculation of angular functions. That is, the amount corresponding to the magnitude of the angle formed by the gradient G and the normal vector N can be calculated with a much smaller calculation amount of only two multiplications and one addition.

Further, in the conventional edge detection method and edge detection device, unnecessary edges 221a to 221a due to the color difference in the foreground object and the intersection of the background objects shown in FIG.
It is unavoidable to mix 221c, and for example, the edge strength in the direction shifted by 45 ° from the estimated edge direction is attenuated by 29%, and the edge strength in the direction shifted by 60 ° is only attenuated by 50%. On the other hand, according to the present embodiment, as shown in FIG. 6, when the direction selection index s is “4”,
It is possible to obtain 75% attenuation for the edge strength in the direction deviated by 45 ° from the estimated edge direction and 94% attenuation for the edge strength in the direction deviated by 60 °. In addition, an edge 221 that may occur due to the color change in the foreground and the background as described above
Since the directions of a to 221c are angles significantly different from the direction of the contour of the target object, use of the strong directivity as described above makes it possible to almost completely eliminate unnecessary edge contamination.

Further, since the tangent vector T, which is the estimated value of the outline direction, and the gradient G of the input image F _in are obtained for each pixel on the outline, an edge along the outline is obtained at each point on the outline. It can be selectively detected.
For example, in the image 250 shown in FIG.
At locations A and B in the circular object 260 that have different colors,
The normal vector N is in the direction perpendicular to the curve 270, and on the horizontal line caused by the color difference in the circular object 260, the gradient G is in the direction perpendicular to the horizontal line. Therefore, the angle difference between the normal vector N and the gradient G is 90 °, and the edge strength E is “0” as shown in FIG. That is, no edge within the circular object 260 is detected. On the other hand, since the edge representing the contour is along the curve 270, the points A and B are
The gradient G at is in the horizontal direction. This is in the same direction as the normal vector N, so that the edge representing the contour is strongly detected. Therefore, in FIG.
The unnecessary edges 221a and 221 as shown in FIG.
Only the edge that is originally desired can be detected without mixing b and 221c.

Further, the tangent vector T, which is the information for estimating the direction of the edge, is given to the edge strength calculation unit 14 as auxiliary information, and the edge strength calculation unit 14 is made to react strongly in a specific direction. Only unnecessary edges can be detected with a high S / N ratio without mixing unnecessary edges.

Further, since the direction of the edge of the target object at each position can be calculated from the outline of the contour input by the operator, only the necessary edge can be detected without increasing the labor of the operator. be able to.
In the above-described embodiment, the set value of the direction selection index s is set to 2
Although a value between 8 and 8 is used, the accuracy of the edge estimated direction may be further increased and a larger value may be set as the direction selection index s. Thereby, the directivity can be further strengthened.

Further, the value of the direction selection index s may be changed from outside by inputting the value of the direction selection index s from the outside. Normally, it is sufficient that the direction selection index s can be selected in several ways. Therefore, the number of additional bits of the cosine table T _C to which the direction selection index s is input may be about 4 bits. Further, the value of the direction selection index s may be limited to an integer, and the computer may calculate the power as repeated multiplication.

In the above embodiment, the cosine "c
Although the power of “osθ” is used to obtain the edge strength, the output of the inner product computing unit 141 shown in FIG.
It is also possible to calculate f (cos ⁻¹ IP) with P and a distribution function of directivity as f (θ), and use the result to obtain the edge strength. Thus, for example, as shown in FIG. 8, when the angle θ formed by the estimated edge direction and the actual edge direction is equal to or less than the threshold value K, the edge strength is not attenuated and exceeds the threshold value K. In this case, it is possible to obtain directivity that sharply attenuates the edge strength.

More specifically, first, a desired distribution shape is defined, and the value on the horizontal axis of the graph shown in FIG. 8 is converted into a cosine. Next, the attenuation rate for each cosine value, that is, the calculation result of “f (cos ⁻¹ IP)” is stored in the table. This table is used in place of the cosine table T _C used in the above embodiment. Therefore, the distribution function f (θ) can be used to obtain the edge strength.

The distribution function f (θ) is the direction selection index s.
May also be held as a variable. That is, the edge strength may be obtained using the distribution function f (θs).
In this case, the direction selection index s is also input to the table, and the edge strength is obtained while changing the direction selection index s.

Further, in the above-described embodiment, the outline of the contour is represented by a spline curve P in the cubic spline format.
Although (t) is used, it may be expressed in a spline format with different orders. In addition, without using the auxiliary variable t, the implicit site form (implicit form)
The curve expressed in the form may be Q (x, y) = 0.

Further, the outline image drawn by the operator may be thinned and the edge direction may be estimated from the positional relationship between adjacent pixels. In this case, in step S3 shown in FIG. 4, the image of the outline is thinned, and in step S4.1 shown in FIG. 4, a plurality of pixels adjacent to the target pixel in the thinned image are displayed. The direction of the contour at that position is determined from the pixel arrangement.

Further, in step S4 shown in FIG. 4, a plurality of filters, such as a compass operator (comp), which selectively reacts with an edge in a specific direction.
The edge strength may be obtained by selecting the filter having the highest detection ability in the direction based on the estimated direction of the edge from the plurality of filters by using the (ass operator).

Further, in the above-mentioned embodiment, the information indicating the outline of the contour is inputted by the operator.
Edges may be continuously detected from a plurality of temporally continuous images, that is, moving images. In this case, in the process of step S2 of the flowchart shown in FIG. 4, the outline is not captured as the coordinate sequence (xn, yn), but the image immediately before and immediately before the image to be currently processed. The contour in the current processing target image is estimated from the edge detection result obtained from the image and the three pieces of information of the current processing target image.

That is, as shown in FIG. 9, first, pixels (hereinafter, referred to as marks) are arranged at regular intervals on the edges 30 obtained as a result of the edge detection processing for the image F _P immediately before the current image F _C to be processed. Say) 31 ₁ , 31 ₂ , 31 ₃ , ...
Tie Next, the marks 31 ₁ , 31 ₂ , 31 ₃ , ...
, A small block including each mark is defined. For example, with respect to the mark 31 ₁ defines a small block 32 which includes a mark 31 _1. The portion having the highest correlation with the data in the small block thus determined is detected by the block matching process from the image F _C currently being processed.

In this block matching process,
First, for example, the pixel value in the small block 32 and the current processing
Image F of the target_CPixel values in blocks of the same size inside
And the level of correlation is determined by the sum of the differences.
Block 33 with the highest correlation
I do. At this time, the pixel 34 in the block 33 is
Mark 32₁Becomes a pixel corresponding to. mark
31₁Other than mark 31 ₂, 31₃The same applies to ...
In this way, the corresponding pixel is set to the current image F to be processed._CDuring ~
To detect from.

Therefore, in the process of step S3 of the flow chart shown in FIG. 4, the spline curve P (t) is generated using the series of corresponding pixel coordinate sequences obtained as described above. As described above, the contour is continuously extracted from the current image to be processed and the immediately preceding image by the block matching process, and the edge is detected using the contour, so that the moving image is automatically processed without operator intervention. Edge can be detected. Also,
In this case, the image used for the block matching process is not limited to the current image to be processed and the image immediately before, and the block matching process may be performed by increasing the number of images.
As a result, the accuracy of calculating the block movement amount can be increased, so that the edge can be detected more accurately.

The key frame method may be applied to the process of step S2 of the flowchart shown in FIG. 4 for a plurality of images that are temporally separated. That is, a method may be applied in which the operator gives a shape instruction to a plurality of images and the computer substitutes the operator by interpolation for intermediate images. In this case, in the process of step S2, the operator sets a rough outline, and for an intermediate image group, a point group on the rough outline is interpolated to automatically generate a spline curve in the image. .

As described above, the edges in the image to be currently processed are detected on the basis of the plurality of images and the results of the edge detection in the past, so that the edge direction or the color can be more accurately used by using the correlation between the frames. It is possible to obtain auxiliary information that is information on changes in the. Therefore, the edge detection accuracy can be further improved, and the edge detection in the moving image can be performed without trouble.

[0089]

According to the edge detection method of the present invention, the edge direction of each position in the input image data composed of grayscale data is estimated. Also, the gradient at each position is obtained. Then, the estimated edge direction information is used as auxiliary information, edge strength is calculated from the auxiliary information and the gradient, and a pixel group in which the pixel value changes sharply compared with the surroundings is detected as an edge. This allows
In the edge detection, strong directivity can be obtained, and only the edges forming the contour of the target object can be detected without mixing unnecessary edges. Therefore, the edge detection accuracy can be improved.

Further, in the edge detection method according to the present invention,
The edge strength is calculated by obtaining the cosine of the angle formed by the estimated edge direction and the gradient. Thereby, the edge strength can be calculated with a small calculation amount. Therefore, a good edge image can be easily obtained.

Further, in the edge detection method according to the present invention,
An angle θ formed by the estimated edge direction and the gradient, a proportional constant C, and a magnitude G of the gradient.
And the index S that determines the directivity, the edge strength E is calculated by the calculation of E = C | G | cos ^S (θ). Thereby, the edge strength can be calculated with a small calculation amount. Therefore, a good edge image can be easily obtained.

Further, in the edge detection method according to the present invention,
The cosine of the angle formed by the estimated edge direction and the gradient is obtained as the inner product of the estimated edge direction, which is a previously normalized vector, and the gradient whose length is normalized to "1". As a result, stronger directivity can be obtained, and the edge strength can be calculated with a smaller amount of calculation. Therefore, a good edge image can be obtained more easily.

Further, in the edge detecting method according to the present invention,
The edge strength is calculated from the auxiliary information and the gradient using the edge strength information table calculated in advance. As a result, the calculation process of the edge strength can be simplified, and a good edge image can be obtained at high speed.

In the edge detecting device according to the present invention, the estimating means estimates the edge direction of each position in the input image data composed of grayscale data, and uses the information of the estimated direction as auxiliary information. The gradient calculating means calculates the gradient at each position in the input image data. The edge strength calculating means calculates the edge strength from the auxiliary information obtained by the estimating means and the gradient obtained by the gradient calculating means. The detection unit detects, as an edge, a pixel group in which the pixel value changes sharply compared with the surroundings, from the edge strength obtained by the edge strength calculation unit. Thereby, strong directivity can be obtained in edge detection, and only edges forming the contour of the target object can be detected without mixing unnecessary edges. Therefore, the edge detection accuracy can be improved.

Further, in the edge detecting device according to the present invention,
The edge strength calculating means calculates the edge strength by obtaining the cosine of the angle formed by the estimated edge direction and the gradient. Thereby, the edge strength can be calculated with a small calculation amount. Therefore, a good edge image can be easily obtained.

Further, in the edge detecting device according to the present invention,
The edge strength calculating means has an angle θ formed by the estimated edge direction and the gradient, a proportional constant C, a magnitude G of the gradient, and an index S for determining directivity, and an edge strength E is E = C | G | cos ^S (θ). Thereby, the edge strength can be calculated with a small calculation amount. Therefore, a good edge image can be easily obtained.

Further, in the edge detecting device according to the present invention,
The edge strength calculation means calculates the cosine of the angle formed by the estimated edge direction and the gradient from the estimated edge direction, which is a vector that is normalized in advance, and the gradient whose length is normalized to “1”. Calculate as the dot product. As a result, stronger directivity can be obtained, and the edge strength can be calculated with a smaller amount of calculation. Therefore, a good edge image can be obtained more easily.

Further, in the edge detecting device according to the present invention,
The edge strength calculation means calculates the edge strength from the auxiliary information and the gradient using an edge strength information table calculated in advance. As a result, the calculation process of the edge strength can be simplified, and a good edge image can be obtained at high speed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an edge detection device according to the present invention.

FIG. 2 is a block diagram showing a configuration of a gradient calculation unit of the edge detection device.

FIG. 3 is a block diagram showing a configuration of an edge strength calculation unit of the edge detection device.

FIG. 4 is a flowchart showing an edge detection process in the edge detection device.

FIG. 5 is a diagram for explaining a curve that gives a rough contour.

FIG. 6 is a graph showing a relationship between a direction selection index and edge strength.

FIG. 7 is a diagram for explaining a relationship between an input image and an edge strength image.

FIG. 8 is a diagram for explaining a case where an edge strength is obtained using a directivity distribution function.

FIG. 9 is a diagram for explaining a block matching process.

FIG. 10 is a schematic diagram showing a Sobel filter for detecting edges in a specific direction.

FIG. 11 is a diagram for explaining a result when edge detection is performed on an image having a color difference inside an object using a conventional edge detection method.

FIG. 12 is a diagram for explaining results when edge detection is performed on an image having a color difference inside an object using a conventional edge detection method that uses an outline of a contour input by an operator.

FIG. 13 is a schematic diagram showing a Robinson 3-level filter group that detects edges in a specific direction.

FIG. 14 is a diagram for explaining a comparison between directivity in a conventional edge detection method and directivity required for practical use.

[Explanation of symbols]

 10 Edge Detection Device 11 Gradient Calculation Unit 12 Spline Curve Generation Unit 13 Parameter Coordinate Generation Unit 14 Edge Strength Calculation Unit

Claims (10)

[Claims] 1. An edge detection method for detecting, as an edge, a pixel group in which pixel values change sharply from the surroundings in input image data composed of grayscale data. While estimating the edge direction of the position, the gradient at each position is obtained, the information of the estimated edge direction is used as auxiliary information, and the edge is detected by calculating the edge strength from the auxiliary information and the gradient. Detection method. 2. The edge detection method according to claim 1, wherein the edge strength is calculated by obtaining a cosine of an angle formed by the estimated edge direction and the gradient. 3. An angle θ formed by the estimated edge direction and the gradient, a proportional constant C, a magnitude G of the gradient, and an index S for determining directivity,
The edge detection method according to claim 1, wherein the edge strength E is calculated by a calculation of E = C | G | cos ^S (θ). 4. The inner product of the estimated edge direction, which is a vector obtained by normalizing the cosine of the angle formed by the estimated edge direction and the gradient, and the gradient whose length is normalized to “1”. The edge detecting method according to claim 1, wherein the edge detecting method is obtained. 5. The edge detecting method according to claim 1, wherein the edge strength is calculated from the auxiliary information and the gradient by using a previously calculated edge strength information table. 6. An edge detection device for detecting, as an edge, a pixel group in which pixel values are sharply changed from the surroundings, from input image data composed of grayscale data, Estimate the edge direction of the position,
From the gradient obtained by the estimating means that uses the estimated direction information as auxiliary information, the gradient calculating means that obtains the gradient at each position in the input image data, the auxiliary information obtained by the estimating means, and the gradient obtained by the gradient calculating means. An edge detecting apparatus comprising: an edge strength calculating means for calculating an edge strength; and a detecting means for detecting an edge from the edge strength obtained by the edge strength calculating means. 7. The edge detecting apparatus according to claim 6, wherein the edge strength calculating means calculates the edge strength by obtaining a cosine of an angle formed by the estimated edge direction and the gradient. 8. The edge strength calculating means has an angle θ formed by the estimated edge direction and the gradient, a proportional constant C, a magnitude G of the gradient, and an index S for determining directivity, 7. The edge detecting device according to claim 6, wherein the edge strength E is calculated by a calculation of E = C | G | cos ^S (θ). 9. The edge strength calculation means normalizes a cosine of an angle formed by the estimated edge direction and the gradient to the estimated edge direction which is a vector that is normalized in advance, and normalizes the length to “1”. 7. The edge detection device according to claim 6, wherein the edge product is obtained as an inner product with the gradient. 10. The edge strength information table calculated in advance, wherein the edge strength calculation means calculates the edge strength from the auxiliary information and the gradient using the information table. Edge detector.