JP2007188407A - Image processing device and image processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To stably detect an eyebrow part from a person's face image. <P>SOLUTION: The image processing device 10 comprises: a reference characteristic point detecting part 14 for detecting characteristic points other than the characteristic point of an eyebrow; a search range setting part 16 for determining the search range of the characteristic points to be detected based on the detected reference characteristic points; a separability map creating part 18 for calculating the output of a separability filter estimated from the positional relationships of the reference characteristic points; a region dividing part 20 for dividing the separability map into a plurality of connected regions; and a region determination part 22 for determining whether or not each region includes a characteristic point. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image processing apparatus and an image processing program.

  As a conventional technique for detecting eyebrows, as disclosed in Patent Document 1, there is a technique in which a black portion is extracted as eyebrows by binarizing an image in a search area. However, depending on illumination conditions and image quality, There was a problem that the eyebrows could not be detected.

  Even if the image is good, there is a problem that it cannot be detected correctly when the eyebrows are thin by an individual or when the color of the eyebrows is not black.

Further, as a technique for detecting facial feature points, as disclosed in Patent Document 2, there is a technique for detecting feature points such as pupils and nostrils using a circular separability filter. Has a problem that it is limited to those showing a circular shape or a shape close thereto.
JP 2000-31124 A Japanese Patent No. 3279913

  As described above, the conventional technique has a problem that it cannot be detected correctly when the eyebrows are thin or the color of the eyebrows is not black. In addition, there is a problem that only a circular shape or a shape close thereto can be detected.

  Accordingly, the present invention has been made to solve the above-described problems of the prior art, and an object thereof is to provide an image processing apparatus and an image processing program for stably detecting a feature point or region of a person. .

  The present invention provides a reference feature for detecting, as a reference feature point, a detection feature point to be detected or a feature point other than the detection region in an image processing apparatus that detects a feature point or region of a human face in an image. A point detection unit, a search range setting unit for determining a search range of the detection feature point or the detection region based on the position of the reference feature point, and a region within the search range, the separability value of the separability filter A separability map creating means for creating a separability map using the image, an area dividing means for dividing the separability map into a plurality of connected areas based on the separability values, and each connected area in the detection area And an area determination unit that determines whether each connected area has a predetermined condition as to whether or not the area includes the detected feature point. apparatus A.

  According to the present invention, in the image processing apparatus that detects a feature point or region of a person's face in an image, the detection feature point to be detected or a feature point other than the detection region is detected as a reference feature point. One or more separation of a reference feature point detection unit, a search range setting unit that determines a search range of the detection feature point or the detection region based on the position of the reference feature point, and a region within the search range A separability map creating means for creating a separability map using a separability value of the degree filter; an area dividing means for dividing the separability map into a plurality of connected areas based on the separability value; A region determination means for determining whether each of the connected regions has a predetermined condition as to whether the region is the detection region or a region including the detection feature point; and the detection feature The point is The eyebrow end point of the object or the detection area is an eyebrow area, and the reference feature point is at least two or more of the left and right pupil centers of the person or the eye end, and the end point of the connection area Whether the end point is an end point of the eyebrow of the person from at least one of an angle and a distance between the position of the end point and the position of each reference feature point. It is an image processing apparatus characterized by determining whether or not.

  According to the present invention, it is possible to stably detect facial feature points and regions of a person.

  Hereinafter, an image processing apparatus 10 according to an embodiment of the present invention will be described with reference to the drawings.

(First embodiment)
An image processing apparatus 10 according to the first embodiment will be described with reference to FIGS. In the present embodiment, an example in which an end point inside the eyebrows is detected is shown.

(1) Configuration of Image Processing Device 10 FIG. 1 is a block diagram showing the image processing device 10.

  An image processing apparatus 10 includes an image input unit 12 that inputs an image, a reference feature point detection unit 14 that detects a feature point or a feature point other than a region to be detected, and a predetermined method based on the detected reference feature point A search range setting unit 16 that determines a search range of a feature point or region to be detected in step S1, and a separability map that calculates the output of a separability filter of one or more sizes estimated from the positional relationship of the reference feature points A creating unit 18; a region dividing unit 20 that divides the separability map into a plurality of connected regions; a region determining unit 22 that determines whether each region is a detection target or a region including feature points; and a result Is output.

  The function of each part 12-22 is realizable with the program memorize | stored in the computer.

(2) Operation of Image Processing Device 10 Next, the operation of the image processing device 10 will be described with reference to FIG.

(2-1) Image input unit 12
First, the image input unit 12 inputs a target image.

(2-2) Reference feature point detector 14
Next, the reference feature point detection unit 14 detects a reference feature point. In the present embodiment, the center point of the left and right pupils is detected.

  As a method for detecting the pupil, for example, a method combining a resolution filter and pattern matching disclosed in Patent Document 2 is used.

(2-3) Search range setting unit 16
The search range setting unit 16 sets a search range based on the reference feature points detected by the reference feature point detection unit 14. FIG. 2 is a diagram showing the set search range.

  As shown in FIG. 2, the search range is determined using parameters determined based on the statistically acquired distribution of the position of the inner edge of the eyebrow, based on the position of the reference feature point (here, the pupil). To do. First, assume that the straight line connecting both pupils is horizontal.

  FIG. 3 is a diagram showing a detailed search range determination method. Here, the left eyebrow will be described as an example, but the same applies to the right eyebrow. As shown in FIG. 3, a rectangular search area 302 expressed by parameters xmin, ymin, xmax, and ymax is set as the search range, with the left pupil center 301 as a reference.

  Each parameter is normalized. For example, when normalized by the distance between both pupils, a value obtained by multiplying the distance between both pupils by the normalized parameter is used.

  Since the straight line connecting both pupils is not necessarily horizontal, the angle θ between the straight line connecting both pupils and the horizontal line is calculated, and the image is rotated by an angle −θ around the midpoint of the line segment connecting both pupils. Is used. In the future processing, this rotated space will be processed only within the range of the search area, and the result obtained at the end is rotated by an angle θ around the midpoint of both pupils. Return to the coordinates of.

(2-4) Separability map creation unit 18
The separability map creating unit 18 creates the separability map by calculating the output value of the circular separability filter for each pixel i of the image in the search area. The degree of separation value η is an index representing the degree of separation between two regions, and is a value calculated by Equation 1.

  Usually, the luminance value of each pixel i in the monochrome image is used as the luminance Pi.

  Regions 1 and 2 create a separation map of two regions as the outer side (region 1 in FIG. 4) and the inner side (region 2 in FIG. 4) of the circle of radius r in the circular separability filter shown in FIG. To do.

  The separability filter having such a shape can extract an object having a circular shape, but it can also extract an elongated object having the same width as twice the radius r. It is.

  The radius r, which is a parameter representing the scale, is calculated by multiplying a predetermined coefficient with reference to the distance between both pupils. Specifically, since the standard face size is known from the distance between the pupils, and the standard eyebrow thickness is known from the face size, the coefficient is calculated from this thickness.

An example of the separability map is as indicated by 502 in FIG. Further, the eyebrows are often darker than the skin color region that is usually the background. Therefore, as in Equation 2, the difference between the average values of the two areas of the separation degree filter may be added to the separation degree value η to obtain the separation degree value η ′. Where α is a predetermined constant.

  Further, the radius r need not be single, and a plurality of radiuses may be used. In that case, the same number of separability maps will be created.

(2-5) Region dividing unit 20
Next, the region dividing unit 20 first binarizes the separability map obtained by the separability map creating unit 18. Binarization sets the value to 1 if the separability value of each point in the separability map is equal to or greater than a predetermined threshold, and to 0 otherwise. An example of the binarized image is shown at 503 in FIG.

  The binary value may be binarized by providing a threshold value not only for the separability value but also for the difference between the average values of the two areas of the separability filter. The other values are set to 0. Here, the value is binarized to 0 or 1 for convenience, but this value itself is not limited to this.

  Next, a labeling process is performed on the separability map. The labeling process is a process of dividing a pixel having a value in the separability map into a plurality of connected regions, assuming that pixels having a value (non-zero) are connected, belonging to the same label. For example, the following four-neighbor method is used.

  First, a pixel to which no label is added is found on the image, and a new label is added.

  Next, the same label is added to the pixels connected to the pixels in four directions. This operation is repeated as long as there is an image to add a label in the image.

  As a result, the separation degree map is divided into a connected region having a binarized value of “0” and a connected region having a binarized value of “1”.

(2-6) Region determination unit 22
Next, the area determination unit 22 determines whether each of the connected areas divided by the area dividing unit 20 includes a target feature point (here, the inner edge of the eyebrows). That is, the maximum and minimum values of the labeled coordinate values are obtained, and the end points are obtained.

  The determination is made on the condition that each connected area touches the right end of the search area and does not touch the upper end, the lower end, and the left end.

  Here, out of candidates that satisfy this condition, the one with the leftmost point of the connected region closest to the predetermined reference point is selected. It is determined based on the distribution of the eyebrow inner edge position statistically acquired in advance.

  Up to this point, the case of the inner edge of the left eyebrow has been described, but in the case of the right eye, the same processing is performed with the left and right reversed.

(2-7) Output unit 24
The output unit 24 outputs the position of the detected end point of the eyebrow.

(3) Effect As described above, according to the image processing apparatus 10 according to the first embodiment, by extracting a region having a certain width by using the circular separability filter, it is possible to reduce the illumination condition and the like. However, the eyebrow end point can be detected stably.

(Second Embodiment)
Next, a second embodiment of the present invention will be described with reference to FIG.

  FIG. 6 is a block diagram showing an image processing apparatus 10 according to the second embodiment. In the present embodiment, a case where a feature point (inner eyebrow end) is not detected, or a case where the detected right and left eyebrow inner ends do not satisfy a predetermined condition will be described.

  The configuration of the image processing apparatus 10 of the present embodiment is obtained by adding an end point estimation unit 26 to the image processing apparatus 10 of the first embodiment, and the other configuration is the same as that of the first embodiment.

  When there is no region that satisfies the predetermined condition in the region determination unit 22, the end point estimation unit 26 uses the reference point used in the region determination unit 22 instead of the detected end point.

  However, in the case of the points estimated in this way, the reliability is assumed to be lower than when there is an area that actually satisfies the condition.

  Further, with respect to the left and right end points thus detected, an angle between a straight line connecting these end points and a straight line connecting the left and right pupil centers is smaller than a predetermined angle (for example, 10 degrees to 20 degrees). If the above condition is satisfied, both end points are detected points.

  When the conditions are not met, if there is a difference in the reliability described above on the left and right, use the higher reliability, and for the lower reliability, search for the point with the higher reliability. In the figure, a position that is symmetrical with respect to the low reliability side is taken as a detection result. If there is no difference in reliability, the one closer to the reference point is trusted, and a position that is symmetrical on the opposite side is set as the detection result as in the case where there is a difference in reliability.

(Example of change)
Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

  In the above embodiment, the left and right pupil centers are used as the reference feature points, but the present invention is not limited to this. For example, each of the end points of the eyes such as the corners of the eyes and the eyes, or the midpoints of the corners of the eyes and the like can be used as the reference feature points.

  Moreover, although the method of labeling the binarization separation map has been described as the region dividing method, the method is not limited to this, and for example, a method using distance conversion, a method for obtaining a ridge line, and the like are also conceivable.

  Moreover, although the example which calculates | requires an end point was described, the area | region itself may be output, the end point of the outer side of an eyebrow may be detected, and not only an eyebrow end but an eye end, a mouth end, etc. are suitable. By setting the search range, detection is possible in the same manner.

  Further, not only the circular separability filter in the above embodiment, but also an elliptical or polygonal separability filter may be used.

1 is a block diagram illustrating a configuration of an image processing apparatus 10 according to a first embodiment of the present invention. It is a figure which shows the set search range. It is a figure which shows the detail of the set search range. It is a figure which shows the example of a separability filter. It is a figure which shows the example of a process result. It is a block diagram which shows the structure of the image processing apparatus 10 concerning 2nd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Image processing apparatus 12 Image input part 14 Reference | standard feature point detection part 16 Search range setting part 18 Separation degree map creation part 20 Area division part 22 Area determination part 24 Output part

Claims (11)

In an image processing apparatus for detecting feature points or areas of a person's face in an image,
Reference feature point detection means for detecting the detection feature point to be detected or a feature point other than the detection region as a reference feature point;
Search range setting means for determining a search range of the detection feature point or the detection region based on the position of the reference feature point;
A separability map creating means for creating a separability map using the separability values of one or a plurality of separability filters for the area within the search range;
Area dividing means for dividing the separability map into a plurality of connected areas based on the separability value;
An area determination means for determining whether each of the connected areas has a predetermined condition as to whether or not each of the connected areas is the detection area or an area including the detected feature points;
An image processing apparatus comprising: In an image processing apparatus for detecting feature points or areas of a person's face in an image,
Reference feature point detection means for detecting the detection feature point to be detected or a feature point other than the detection region as a reference feature point;
Search range setting means for determining a search range of the detection feature point or the detection region based on the position of the reference feature point;
A separability map creating means for creating a separability map using the separability values of one or a plurality of separability filters for the area within the search range;
Area dividing means for dividing the separability map into a plurality of connected areas based on the separability value;
An area determination means for determining whether each of the connected areas has a predetermined condition as to whether or not each of the connected areas is the detection area or an area including the detected feature points;
The detection feature point is an end point of a person's eyebrow, or the detection region is an eyebrow region, and the reference feature point is at least two or more of the left and right pupil centers of the person, or the eye edge, Endpoint detection means for detecting an endpoint of the connected region;
With
The region determining means determines whether or not the end point is an end point of the eyebrow of the person based on at least one of an angle and a distance between the position of the end point and the position of each reference feature point. Processing equipment. The area determination means includes
The image processing apparatus according to claim 2, wherein when there is no connected area that satisfies the predetermined condition, it is determined whether or not each connected area satisfies the second condition. The image processing apparatus according to claim 2, wherein the separability filter is circular, elliptical, or polygonal. 5. The at least one of claims 2 to 4, wherein the separability value of the separability filter is a value obtained by dividing the sum of inter-class variances of luminance values of two regions by the sum of inter-class variances. Image processing device. 6. The image processing apparatus according to claim 2, wherein the separation degree value of the separation degree filter is a difference value between average values of luminance values in two regions. In an image processing program for detecting a feature point or region of a human face in an image by a computer,
A reference feature point detection function for detecting the detection feature point to be detected or a feature point other than the detection region as a reference feature point;
A search range setting function for determining a search range of the detection feature point or the detection region based on the position of the reference feature point;
A separability map creating function for creating a separability map using the separability value of the separability filter for the area within the search range;
A region dividing function for dividing the separation map into a plurality of connected regions based on the separation value;
An area determination function for determining whether each of the connected areas has a predetermined condition as to whether or not each of the connected areas is the detection area or an area including the detection feature point;
The detection feature point is an end point of a person's eyebrow, or the detection region is an eyebrow region, and the reference feature point is at least two or more of the left and right pupil centers of the person, or the eye edge, An endpoint detection function for detecting an endpoint of the connected region;
Realized,
The area determination function determines whether the end point is an end point of the person's eyebrows from at least one of an angle and a distance between the position of the end point and the position of each reference feature point. Processing program. The area determination function is
The image processing program according to claim 7, wherein when there is no connected area satisfying the predetermined condition, it is determined whether or not each connected area satisfies the second condition. The image processing program according to claim 7 or 8, wherein the separability filter is circular, elliptical, or polygonal. 10. The at least one of claims 7 to 9, wherein the separability value of the separability filter is a value obtained by dividing the sum of inter-class variances of luminance values of two regions by the sum of inter-class variances. Image processing program. The image processing program according to at least one of claims 7 to 10, wherein the separability value of the separability filter is a difference value between average values of luminance values in two regions.