JP3499305B2 - Face area extraction method and exposure amount determination method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a face area extracting method and an exposure amount determining method, and more particularly to a face area determining method for extracting an area corresponding to a human face existing in an original image, and the face area determining method.
The present invention relates to an exposure amount determination method for determining an exposure amount so that the area extracted by the area determination method is printed in an appropriate color.

[0002]

2. Description of the Related Art The most noticeable part of a person's photograph is a person's face. For example, when an original image recorded on a film is printed on photographic paper or the like. Needs to determine the exposure amount so that the color of the person's face is printed in an appropriate color.

Therefore, the applicant of the present invention, based on the rule of thumb that a person is highly likely to be located in the approximate center of the image in a portrait photograph, the film image is divided into a plurality of areas fixedly fixed in advance. Each area is divided and weighted so that the weight of the area located substantially in the center of the image is heavy, the weighted average value of the densities of the three colors of each area is obtained, and the exposure is performed based on the weighted average value. A method for determining the amount has already been proposed (see JP-A-63-80242). However, in the above, if the person is actually located near the center of the image, it is possible to obtain an exposure amount that allows the person to be printed with an appropriate color, but the person is greatly deviated from the center of the image. There is a problem that it is impossible to obtain an appropriate exposure amount when it is positioned.

Further, the applicant of the present invention divides a color original image into a large number of pixels, decomposes each pixel into three colors, and performs photometry, and based on the data obtained by photometry, the hue value (and saturation value) is determined. Obtain the histogram, divide the obtained histogram for each mountain, determine which of the divided mountains each pixel belongs to, divide each pixel into groups corresponding to the divided mountains, and obtain the color original image for each group. It is proposed to divide the image into a plurality of regions, estimate a region corresponding to a person's face in the plurality of regions, and determine the exposure amount based on the photometric data of the estimated region (Japanese Patent Laid-Open No. 4-346332). (See the official gazette).

Further, in Japanese Patent Laid-Open No. 6-160993, in order to improve the accuracy of extracting a region corresponding to a person's face,
The area that touches the outer edge of the image is judged to be the background area and removed, or the extracted area is made into a line figure, and based on the shape of the neighboring area and the shape of the extracted area located around the extracted area. It is also described that it is determined whether or not the extracted region is a region corresponding to a person's face.

However, in the above, the original image has a flesh color area such as the ground or a trunk of a tree, and the hue and saturation of the flesh color area correspond to the hue of the area of the person in the original image. If it is close to the saturation, there is a possibility that this area may be erroneously determined as an area corresponding to a human face. Also, if this skin color area is adjacent to the area corresponding to the person's face, the skin color area and the area corresponding to the person's face may not be separated, and the original image may not be divided into areas within an appropriate range. It was In the above-described conventional technique, the processing is performed on the assumption that one of the divided areas is an area corresponding to a person's face. Therefore, if the original image cannot be divided into areas in an appropriate range, There is a problem in that the area corresponding to the face is erroneously determined, and an exposure amount with which the face of the person can be properly printed cannot be obtained.

The present invention has been made in consideration of the above facts, and an area corresponding to a person's face is not affected by colors of an area corresponding to the person's face in the original image and surrounding areas. Get a face area extraction method that can extract only
That is the first purpose. In addition, the present invention can be applied to the person in the original image.
The influence of the color of the area corresponding to the face of the object and the surrounding area
A second object is to obtain an exposure amount determination method capable of determining the exposure amount so that an area corresponding to a person's face can be properly printed without receiving it .

[0008]

In order to achieve the above object, the face area extracting method according to the invention of claim 1 is the first shape peculiar to the first part of the person existing in the original image. After searching the pattern and setting the area estimated to correspond to the human face according to the size and orientation of the detected first shape pattern and the positional relationship between the human face and the first portion, A second shape pattern unique to at least one second portion different from the first portion of the person is the size and orientation of the set region in the original image, the face of the person, and the second portion. by searching whether present within the search range set in accordance with the positional relationship between the region a high probability a region corresponding to a person's face in the original image you judgment.

According to a second aspect of the present invention, in the first aspect of the invention, the first shape pattern and the second shape pattern represent a contour pattern of a person's head and a contour of a person's face. It is characterized by including at least one of a shape pattern, a shape pattern representing the internal structure of a person's face, and a shape pattern representing the outline of the body of the person.

According to a third aspect of the present invention, in the first aspect of the invention, the first shape pattern and the second shape pattern are searched for by a plurality of pixels having the same or similar density or color to the original image. It is characterized in that it is divided into a plurality of regions constituted by and each contour of each of the plurality of regions is used.

According to a fourth aspect of the invention, in the first aspect of the invention, the first shape pattern and the second shape pattern are searched for by detecting an edge existing in the original image and detecting the center of the detected edge. The feature is that the line is obtained and the obtained center line is used.

According to a fifth aspect of the invention, in the first aspect of the invention, prior to the search for the first shape pattern, an area of low accuracy corresponding to a person in the original image is searched for the first shape pattern. It is characterized by being excluded from. Contract
The invention according to claim 6 is the same as the invention according to claim 1,
As the first shape pattern or the second shape pattern
Of the shape patterns that represent the facial structure of a person.
According to the invention of claim 3, the search for the represented shape pattern is performed.
Divided area or obtained by the invention according to claim 4
Ellipse of the area divided by the centerline of the edge
Area is extracted, and each of the extracted elliptical areas
Similarity over a certain value by matching process using feature amount
Detects a pair of elliptical areas in degrees and finds the pair of elliptic area pairs
Depending on the degree of symmetry, the pair of elliptical regions are the eye parts of the human face.
It is special to do by judging whether it corresponds to the pair.
It is a sign. Further, in the exposure amount determination method according to the invention described in claim 7, the face area extraction method according to any one of claims 1 to 6 is used to determine the probability that the area corresponds to a person's face in the original image. Is determined, and the exposure amount to the copy material is determined based on at least one of the color and the density of the determined region.

[0013]

According to the first aspect of the invention, the first shape pattern peculiar to the first part of the person existing in the original image is searched for, and the size, direction and person of the detected first shape pattern are detected. A region estimated to correspond to a person's face is set according to the positional relationship between the face and the first portion. The first shape pattern may be plural or singular. For example, as described in claim 2, a shape pattern representing a contour of a person's head and a shape representing a contour of a person's face. At least one of a pattern, a shape pattern representing the internal structure of a person's face, and a shape pattern representing the outline of the body of a person can be used. In particular, considering that any of the shape patterns of each part of the person may not be detected depending on the original image, a plurality of shape patterns are set as the first shape pattern, and each of them is searched. Good.

For example, a region on the original image corresponding to a person's head (hair portion) has a contour with a convex portion having a predetermined degree of curvature corresponding to the top of the person's head and a person below the convex portion. Generally, a concave portion having a predetermined circularity corresponding to the boundary between the head and the face is included. Therefore, for a person's head, for example, a convex portion having a circularity within a predetermined range and a concave portion having a circularity within a predetermined range may be used as a specific shape pattern representing the contour of the head. it can. Then, when the above-mentioned shape pattern is used as the first shape pattern and the above-mentioned shape pattern is detected in the original image, an area estimated to correspond to a person's face is extracted, for example, as follows. You can

That is, the face of the person is present at a position adjacent to the lower side (recess side of the above-mentioned shape pattern) with respect to the head, and the size thereof is substantially proportional to the size of the head, and further oriented. Matches the orientation of the head. In addition, the outline of a person's face is generally a substantially elliptical shape. Therefore, when the above-described shape pattern is detected in the original image, the size and orientation of the detected shape pattern are determined according to the positional relationship between the face and the head of the person. It is possible to set an elliptical area at different positions (positions adjacent to the recessed side of the detected shape pattern) and set the area as an area estimated to correspond to a human face.

As described above, in the extraction process according to the first aspect of the present invention, the size and orientation of the first shape pattern peculiar to the first part of the person detected from the original image, and the face and the first face of the person. The area estimated to correspond to the person's face is set according to the positional relationship with the portion of the person's face, so that the person's face is not affected by the color of the area corresponding to the person's face and the surrounding area. The area estimated to be equivalent can be set appropriately. On the contrary, even if the original image includes a skin-colored area having a contour similar to the face area, although the area does not correspond to the person's face, it is prevented from being erroneously determined as an area corresponding to the person's face. it can.

If the area set by using the first shape pattern as described above matches the area corresponding to the face of the person, the first area of the person near the set area.
There is a high possibility that a shape pattern peculiar to another portion different from the above portion exists. Therefore, in the invention according to claim 1, after setting the region estimated to correspond to the face of the person, the second shape pattern peculiar to at least one second portion different from the first portion of the person, The person in the original image is searched by searching for the size and orientation of the set area in the original image and whether or not it exists within the search range set according to the positional relationship between the person's face and the second portion. Area having a high degree of certainty, which is an area corresponding to the face.

The second shape pattern may be plural or singular, and as described in claim 2, the shape pattern representing the contour of the person's head and the contour of the person's face. It is possible to use at least one of a shape pattern that represents a human body, a shape pattern that represents the internal structure of a person's face, and a shape pattern that represents the outline of the body of a person.

As described above, the second shape pattern is detected within the search range set according to the size and orientation of the set area in the original image and the positional relationship between the person's face and the second portion. In this case, it can be determined that the set area is highly likely to be an area corresponding to a person's face. in this way,
For a region that is estimated to correspond to a human face set using the first shape pattern, a search is made as to whether or not the second shape pattern exists within the search range corresponding to that region.
Even if the first shape pattern is accidentally detected in the non-human area in the original image, it is possible to prevent the area determined based on the accidentally detected shape pattern from being erroneously determined as an area corresponding to a human face. It Further, even when a plurality of regions estimated to correspond to the person's face are set by the first shape pattern, it is possible to determine the region having the highest accuracy corresponding to the person's face by the second shape pattern.

Therefore, according to the present invention , the person in the original image is
Are affected by the colors of the area corresponding to the human face and the surrounding area.
Extract only the area corresponding to the person's face without blurring
be able to. Also, regarding the second shape pattern,
Since the search is performed only within the search range set according to the size and orientation of the area set as the area estimated to correspond to the human face and the positional relationship between the human face and the second portion,
The processing time can be shortened as compared with the case where the entire image is searched for each shape pattern, and a highly accurate area corresponding to a human face can be extracted at high speed.

In the search for the shape pattern, for example, as described in claim 3, the original image is divided into a plurality of regions composed of a plurality of pixels having the same or similar density or color, and the plurality of regions are divided. This can be done using the contour of each of the regions. When the image is divided into a plurality of areas as described above, a plurality of areas having a high probability that the contour includes a shape pattern peculiar to each part of the person such as a boundary between the person and the background can be obtained. Therefore, the shape pattern unique to each part of the person can be efficiently detected from the contours of the plurality of regions.

The original image is divided into a one-dimensional histogram of hue values, or a hue value and a saturation value based on photometric data of three colors for each pixel of the original image, as disclosed in, for example, JP-A-4-346332. Is obtained, the obtained histogram is divided for each mountain, it is determined which of the divided mountains each pixel belongs to, and each pixel is divided into a group corresponding to the divided mountain, and for each group The original image can be divided into multiple areas. Further, the original image may be divided into a plurality of areas by setting a pixel having a large density difference or a large color difference with an adjacent pixel among the pixels forming the original image as a boundary of the area. Further, the original image may be divided into a plurality of regions by performing binarization, or the binarization may be performed a plurality of times by using a plurality of threshold values having different values. You may make it divide | segment into a some area | region by the said combination.

In the present invention, the regions divided as described above are used to detect each shape pattern, and the region estimated to correspond to a human face is set from the original image as described above. The size of the detected first shape pattern,
Perform according to the direction. Therefore, even if some of the plurality of shape patterns cannot be detected from the contours of the plurality of regions, it is possible to set a region estimated to correspond to the face of the person and to divide the plurality of divided regions as in the conventional case. It is not necessary for any of the above to match the area corresponding to the face of the person.

Further, in the search for the shape pattern, as described in claim 4, an edge existing in the original image is detected,
The center line of the detected edge may be obtained, and the obtained center line may be used. The edge can be detected by, for example, differentiating the density or color of each pixel, and the centerline of the edge can be obtained by a known process such as thinning or ridge detection. . Since it is highly probable that the edges existing in the image also include a shape pattern peculiar to each part of the person, the shape pattern peculiar to each part of the person can be efficiently detected as in the invention of claim 3. You can

By the way, there are cases where an image has a region having a characteristic that it can be determined that the probability of being a region corresponding to a person is low. For example, a region in which the ratio of the straight line portion included in the contour is equal to or greater than a predetermined value has a high probability of being a region representing an artifact. Further, the human body is substantially line-symmetric with respect to an imaginary line that divides the left and right of the human body, but a region where the degree of symmetry is a predetermined value or less is unlikely to be a region corresponding to a person. Further, it can be determined that the probability that the region corresponding to a person is low even if the number of irregularities is equal to or larger than a predetermined value. In addition, since a person is generally likely to be located in the substantially central portion of the image, it can be determined that an area having a contact rate with the outer edge of the image of a predetermined value or more is also located in the peripheral portion of the image. The probability of being a region corresponding to is low.

If the density contrast in the area is equal to or less than a predetermined value, the area is likely to represent an object having a smooth surface or less unevenness, and is an area corresponding to a person, particularly a face. Is unlikely. Further, even when the density in the area changes in a predetermined pattern or when the density in the area repeats a predetermined change pattern, it can be determined that the probability of the area corresponding to a person is low. 6. A region with high accuracy that does not correspond to any part of a person in this way is defined as a region with low accuracy that corresponds to a face of a person.
As described above, the search for the first shape pattern is performed by excluding the region with low accuracy corresponding to the person in the original image from the search target of the first shape pattern prior to the search for the first shape pattern. The time required for the processing is shortened, and the processing can be performed at higher speed. Further, according to the invention of claim 1,
And the first shape pattern or the second shape pattern
Of the shape patterns that represent the facial structure of a person.
A search for a shape pattern representing
As described above, the area divided by the invention according to claim 3,
Alternatively, the center of the edge obtained by the invention of claim 4.
Extract elliptical areas from the areas divided by lines
Then, each feature amount (shape, shape,
Similarity over a specified value by matching processing using (density)
Detects a pair of elliptical areas in degrees and finds the pair of elliptic area pairs
Depending on the degree of symmetry, the pair of elliptical regions are the eye parts of the human face.
It can be done by judging whether or not it corresponds to the pair.
Wear. An exposure amount determining method according to a seventh aspect of the invention is an area corresponding to the face of a person in the original image, which is determined by the face area extracting method according to any one of the first to sixth aspects. Since the exposure amount to the copy material is determined based on at least one of the color and the density of the highly accurate area, without being affected by the color of the area corresponding to the face of the person in the original image and the area around the area, The exposure amount can be determined so that an area corresponding to a person's face can be properly printed.

[0027]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 shows a photographic printing apparatus 10 to which the present invention can be applied. The photographic printing apparatus 10 includes a conveyance roller 14 that conveys the negative film 12. A light source 16, a color correction filter 18 such as a light control filter, and a diffusion box 20 are sequentially arranged below the conveyance path of the negative film 12. Further, a distribution prism 22 that distributes the light transmitted through the negative film 12 in two directions is arranged above the conveyance path of the negative film 12.

A projection optical system 24, a black shutter 26, and a color paper (printing paper) 28 as a copying material are sequentially arranged on one optical path of the light divided by the distribution prism 22 in two directions. The projection optical system 30 and the CCD image sensor 32 are sequentially arranged on the other optical path. The CCD image sensor 32 divides the entire image (one frame) recorded on the negative film 12 into a large number of pixels (for example, 256 × 256 pixels), and each pixel is red.
(Red), G (green), and B (blue) are separated into three colors for photometry.

An amplifier 34 for amplifying the signal output from the CCD image sensor 32, an analog-digital (A / D) converter 36, and a sensitivity correction of the CCD image sensor 32 are provided at the signal output end of the CCD image sensor 32. 3x
Three matrix circuits 38 are connected in order. The 3 × 3 matrix circuit 38 is composed of a micro processing unit (MPU), and calculates a face area extraction unit 40 in which a program for realizing a face area extraction process, which will be described later, is stored in the ROM, and an average density of the entire image. Each is connected to the average density calculation unit 42. In addition, the face area extraction unit 40
The average density calculation unit 42 is connected to the exposure amount calculation unit 44. The exposure amount calculation unit 44 is connected to the color correction filter 18 via a driver 46 that drives the color correction filter 18.

Next, the operation of this embodiment will be described. Light source 16
The light emitted from the device passes through the color correction filter 18, the diffusion box 20, and the negative film 12, is distributed by the distribution prism 22, and is transmitted through the projection optical system 30 to the CCD.
It is incident on the image sensor 32. At this time, the black shutter 26 is closed. The CCD image sensor 32 divides the entire one screen into a large number of pixels by the incident light, decomposes each pixel into three colors of R, G, and B to perform photometry, and outputs a photometric data signal. The photometric data signal is amplified by the amplifier 34, converted into a digital signal by the A / D converter 36, and further C by the 3 × 3 matrix circuit 38.
The sensitivity of the CD image sensor 32 is corrected, and the image data is input to the face area extraction unit 40 and the average density calculation unit 42, respectively.

The average density calculator 42 calculates the average density of the entire screen. The face area extraction unit 40 estimates a portion corresponding to the face of a person in one screen as described later in detail, and outputs R, G, B three-color photometric data of the portion.
The exposure amount calculation unit 44 calculates the exposure amount using the photometric data of the three colors output from the person area calculation unit 40 and the average density output from the average density calculation unit 42, and the negative film is calculated with the calculated exposure amount. Twelve images are printed on the color paper 28 so that the color correction filter 1 is passed through the driver 46.
8 is moved and the black shutter 26 is opened / closed to print an image.

The average density of the entire screen calculated by the average density calculation unit 42 is not essential information in the exposure amount calculation by the exposure amount calculation unit 44, and the average density calculation unit 42 is omitted and the exposure amount calculation unit is omitted. At 44, the exposure amount may be calculated using only the photometric data of the three colors output from the face area extraction unit 40.

Next, the face area extraction processing executed by the CPU (not shown) of the face area extraction unit 40 will be described with reference to the flowchart of FIG. In step 100, the image data from the 3 × 3 matrix circuit 38 is fetched.

In step 102, the image data is binarized with a predetermined threshold value as an example of a method of dividing the original image into a plurality of regions. Hereinafter, an example of a positive image will be described. Due to this binarization, the original image has an area composed of pixels each having a density higher than the threshold value (hereinafter, this area is referred to as a “black area”) and each pixel having a density lower than the threshold value. Area composed (hereinafter, this area is referred to as "white area")
Is divided into and As a result, as an example, in the original image shown in FIG. 8A, a large number of black areas including the black area 50 corresponding to the hair portion of the person in the original image is extracted as shown in FIG. 8B. become. Further, in step 102, in order to identify each of the plurality of divided black areas and white areas, numbering is performed on each area.

In step 104, a non-human area determination process is performed. This non-personal area determination process will be described with reference to the flowchart in FIG. In step 130, FIG.
In the step 102 of the flowchart of FIG. 2, the straight line portion of the contour is detected for each of the black region and the white region extracted by binarization, and in step 132, the ratio of the straight line portion in the contour is calculated for each region. . In step 134, it is determined whether or not there is a region where the ratio is equal to or greater than a predetermined value.
It can be determined that a region in which the ratio of the straight line portion occupying the contour is equal to or greater than a predetermined value has a high probability of being a region representing an artifact and a low probability of being a region corresponding to a person. Therefore, if the determination in step 134 is affirmative, step 13
An area in which the ratio of the straight line portion is equal to or greater than a predetermined value in 6 is excluded from the determination target of the area representing the contour of the head of the person described later (thus, the area is excluded from the search target of the first shape pattern). ), And proceeds to step 138.

In step 138, a line symmetry axis extending along the vertical direction of the image is set in the vicinity of the center of the image in the left-right direction in each region, and in step 140, a line relating to the line symmetry axis set above is set. The symmetry similarity is calculated for each area. The method of calculating the line symmetry similarity will be described later. In step 142, it is determined whether or not there is a region whose line symmetry similarity is equal to or lower than a predetermined value. Generally, a human body is substantially line-symmetric with respect to an imaginary line that divides the left and right of the human body, and it can be determined that a region having a line symmetry similarity of a predetermined value or less is unlikely to be a region corresponding to a person. Therefore, if the determination in step 142 is affirmative, step 14
In step 4, the area having the line symmetry similarity less than or equal to the predetermined value is excluded from the determination target of the area representing the contour of the head of the person.
Move to.

In step 146, the contact rate of the contour with the image outer edge is calculated for each area, and in step 148, it is determined whether or not there is an area having a contact rate of a predetermined value or more. Although it can be determined that the area where the contact rate is equal to or higher than the predetermined value is located in the peripheral portion of the image, in general, the probability that the person is located in the substantially central portion of the image is high, and thus the area where the contact rate is equal to or higher than the predetermined value. It can be determined that there is a low probability that the area corresponds to a person. Therefore, when the determination in step 148 is affirmative, the region in which the contact rate with the image outer edge is equal to or more than the predetermined value is excluded from the determination target of the region representing the contour of the head of the person in step 150,
The process proceeds to step 152.

In step 152, the contrast of the density inside each area (difference between the maximum density value and the minimum density value) is calculated, and in step 154, it is determined whether or not there is an area where the density contrast is below a predetermined value. When the contrast of the density in the area is less than or equal to a predetermined value, there is a high possibility that the surface represents an object whose surface is smooth or has few irregularities, and the probability that the area corresponds to a person is low. Therefore, if the determination in step 154 is affirmative, in step 156, the region where the internal density contrast is equal to or less than the predetermined value is excluded from the determination target of the region representing the contour of the head of the person, and step 1
Move to 58.

In step 158, the density value of each pixel in each area is differentiated along a plurality of different directions (for example, the horizontal direction of the image, the vertical direction, and the direction inclined by ± 45 ° with respect to the horizontal direction). In step 160, it is determined whether or not there is a region in which at least one of the differential values along a plurality of directions changes regularly. The above determination is affirmed for the region where the density changes in a predetermined pattern or the region where the density repeats the predetermined change pattern. Since it is possible to determine that such a region has a low probability of being a region corresponding to a person, if the determination in step 160 is positive, the region in which the differential value regularly changes in step 162 is determined to be the person's head. Exclude from the judgment target of the area representing the contour of the part. As described above, the non-personal area determination process is completed, and the process proceeds to step 106 in the flowchart of FIG.

In the next step 106, face candidate area setting processing by head extraction is performed. This processing searches for a shape pattern representing the contour of the head of the person as a shape pattern (first shape pattern) peculiar to the head as the first part of the person, and sets a face candidate area. Hereinafter, description will be made with reference to the flowchart of FIG.

In step 172, one of the black areas extracted by the above-described binarization that has not been excluded by the non-human area exclusion processing described above is selected.
Take out one. In step 174, the contour of the extracted black area is traced and the curvature of the contour is calculated. For example, in FIG.
A large number of black regions are extracted by binarizing the original image shown in FIG. 8A. If a black region 50 as shown in FIG. 8B is taken out, the contour trace is shown in FIG. It can be done clockwise as shown in B).
Further, as shown in FIG. 8C, the curvature is calculated by sequentially setting a vector having a predetermined length, a direction substantially along the tracing direction, and a start point and an end point of which are in contact with the contour of the black region (see FIG. 8). In (C), vector P1P0 and vector P0P2),
The curvature θ can be calculated from the inner product of adjacent vectors. The direction of θ (the sign of the sign) can be obtained from the cross product of the vectors.

At this time, the angle calculation expression is set so that the contour of the black region has a negative value when the contour of the black region curves to the right and a positive value when the contour of the black region curves to the left with respect to the traveling direction of the trace. I shall. For example, FIG. 8C shows a portion of the contour of the black region that curves to the right with respect to the traveling direction of the trace, and θ has a negative value.

The length of the vector may be fixed or may be changed according to the peripheral length of the outline of the black area. Further, the curvature may be calculated a plurality of times, and the length may be changed stepwise at each time.

In the next step 176, based on the curvature of the contour of the black region calculated above, a shape pattern representing the contour of the head of the person corresponds to the contour and the boundary between the head and the face of the person. Then, the black region having the concave portion estimated and the convex portion estimated to correspond to the top of the person's head is extracted.
The concave portion and the convex portion can be determined based on the tracing direction and the previously obtained direction of θ (sign positive / negative). That is, a portion where the sign of θ is reversed is the boundary between the concave portion and the convex portion, and when traced clockwise, it can be determined that the concave portion is a positive portion and the convex portion is a negative θ. .
Thereby, as an example, as shown in FIG. 8D, it is possible to extract the concave portion and the convex portion in the contour of the black region.

At step 178, the respective feature quantities are calculated for the concave portion and the convex portion extracted above. In the present embodiment, the circularity calculated as follows is used as the characteristic amount of the concave portion and the convex portion. That is, as shown in FIG. 9A, the length of the curve forming the concave portion or the convex portion is L, the distance between the end points Q1 and Q2 of the concave portion or the convex portion is M, and the straight line connecting the end points Q1 and Q2. Let h be the height of the concave portion or the convex portion with respect to, and calculate L / M and h / M as the degree of circularity. In addition, a direction vector V is used as a vector toward the opening direction of the concave portion and the convex portion.
Ask for.

In step 180, it is determined whether the black region can be determined to be the person's head (hair portion), that is, the probability of being the person's head is high, using the feature amount calculated above. In this determination, the concave portion and the convex portion of the contour of the predetermined value or more are respectively extracted from the contour of the black area, and the ratio of the length of the concave portion and the convex portion to the peripheral length of the contour of the black area is within the predetermined range, respectively. Further, it is affirmed when the consistency evaluation is higher as the person's head than the position and direction. This consistency evaluation can be performed as follows as an example. First, the conformity in units of concave portions and convex portions is evaluated by the following procedure.

Any three points on the curve of the concave or convex portion (for example, both end points and the center point of the curve, but not on the same straight line) are selected, and a circle passing through each of the selected three points is assumed. , Find the center point of the circle.

The distances between all the pixels that form the concave or convex curve and the center point of the circle are obtained, and the curves whose distance variation is equal to or less than a predetermined value are evaluated as matching. If the variation is larger than a predetermined value, it is excluded from the evaluation target, or the curve is further divided at a portion having a large curvature and the processing is performed again.

Next, the conformity including the concave portion and the convex portion is evaluated as follows. By using the center of gravity of the center point of the circle obtained for each curve of the concave and convex parts, a comprehensive center point and a comprehensive center area (a circular area with a radius determined according to the average of the lengths of the curves) , The area indicated by the broken line in FIG. 9B) is set. When the center point obtained for each curve is within the comprehensive center area, or when the center axis of each curve (direction vector V or its extension line) is within the center area, It is evaluated that the matching as a person's head is high.

Further, instead of the above evaluation, in the same manner as above, when the variation in the distances between all the pixels forming the curved lines of the concave and convex portions and the center point of the circle is less than a predetermined value. In addition, it is evaluated that the matching as a person's head is high.

When the determination in step 180 is affirmative, most of the person has a concave hair portion at the boundary between the hair portion and the face of the person.
In step 182, an elliptical area inscribed in the recess is set as a face candidate area (area estimated to correspond to a person's face) based on the size of the recess of the outline of the black area and the center position thereof. After executing step 182, the process moves to step 186. Also, step 180
If the determination is negative, the process proceeds to step 186 without performing any processing.

In step 186, it is determined whether the above process has been performed for all the black areas extracted by binarization. If the determination in step 186 is negative, the process returns to step 172, and steps 172 to 186 are repeated. When the determination in step 186 is affirmative, the face candidate area setting process by head extraction is ended, and the process proceeds to step 108 in the flowchart of FIG.

In step 108, a matching determination process based on the face contour is performed. This processing searches for a shape pattern that represents the contour of the person's face as a shape pattern (second shape pattern) peculiar to the face as the second part of the person, and determines the probability that the area corresponds to the person's face. Is used to determine a face candidate area having a high score, which will be described below with reference to the flowchart in FIG.

In step 190, one of the face candidate areas set by the face candidate area setting process by head extraction described above is taken out. In the next step 192, the size and orientation of the extracted face candidate area and the face of the person
The search range for searching the shape pattern representing the face contour is set according to the positional relationship with the portion (face contour). As an example, when the extracted face candidate area is in a range surrounded by a broken line in FIG. 10A, a shape pattern representing the contour of the face (here, the side contour of the face) with respect to this face candidate area. The range indicated by hatching in FIG. 10A, which is estimated to exist, is set as the search range.

In the next step 194, among the many areas (black area and white area) extracted by the binarization, the areas existing in the search range are processed in step 174 of the flowchart of FIG. 4 described above. Similarly to 176, the contour of each area is traced, the curvature of the contour is calculated, and the concave portion and the convex portion are extracted. As a result, for example, FIG.
For the original image shown in FIG. 10A, the concave and convex portions existing within the search range shown in FIG.
In addition, a large number of concave portions and convex portions including the concave portions and convex portions indicated by are extracted.

After extracting the concave portions and the convex portions, priority is given to a curve having a long length among the curves extracted as the concave portions or the convex portions so that the curvature (angle θ) falls within a predetermined range. Alternatively, the curve may be further divided so that the degree of curvature is equal to or less than a predetermined value, or the angle of the curve in the normal direction is roughly classified. For example, FIG. 10 (C)
In the case where both the curve and the curve are within the search range and they are extracted as a continuous curve, they are divided into two curves because their circularity is large. In the following, the curves obtained by dividing the contours of the respective areas extracted as described above are collectively referred to as "lines".

In step 196, as the shape pattern representing the contour of the human face, a line estimated to correspond to the side part of the human face is selected from the large number of concave portions and convex portions (lines) extracted in step 194. , I.e., all pairs of lines that can be candidates for face contour are extracted. In particular,
Among the lines extracted above, the direction vector V of each other
Are facing each other, the magnitude of the intersection angle of the direction vectors V (or their extension lines) with each other is within a predetermined value, and the similarity of the length and the degree of circularity is a predetermined value or more. All are extracted as pairs of lines that can be candidates for face contours.
For example, in the example shown in FIG. 10, line-to-line pairs and the like are extracted from the many lines extracted in step 194 as line pairs that can be candidates for the face contour. Note that V2 and V4 are direction vectors of each line.

In the next step 198, it is determined whether or not there is a pair of lines that can be candidates for the face contour. When a line that meets the above-described conditions is extracted from the search range within the search range, the determination at step 198 is affirmative and the process proceeds to step 200.

In step 200, for all the line pairs extracted in step 196, a line symmetry axis is set between the line pairs as shown in FIG.
At 02, the line symmetry similarity of the line pair is determined for each line pair. The determination of the line symmetry similarity is first performed in FIG.
As shown in, the length of both lines along the direction parallel to the axis of line symmetry is v, the interval between both lines along the direction orthogonal to the axis of line symmetry is h, and the aspect ratio v / h is the face of the person. It is determined whether the value is within a predetermined range corresponding to the aspect ratio of. Next, FIG.
As shown in (A), a matching expansion area is set for each line (such as a process of expanding around the line), and a curved area pattern symmetric with respect to the line symmetry axis is obtained for one of the line pairs (in the figure, The line 'which is symmetrical to the line'), the similarity between the obtained curved area pattern and the other line extended area pattern is determined.

As a method of determining the degree of similarity, for example, the cumulative value of the difference absolute value or the squared difference value between the pixel data existing at corresponding positions of the area patterns to be matched is calculated, and this cumulative value is small. It can be determined that the similarity between areas is larger as As the data of each pixel, the image data itself, the binarized data, the data obtained by differentiating the image data, or the like can be used.

Further, in the next step 204, when it is assumed that each line pair is a line corresponding to both sides of a person's face, the direction vector is nearly vertical to the position corresponding to the boundary with the head. Search for the existence of the line. For example, with respect to the line and the pair of lines described above, a range 52A shown in FIG. 11B is searched for as a position corresponding to the boundary with the head, and the direction vector existing in this range 52A becomes vertical. A line will be extracted as a close line.

In the next step 206, each line pair is a line pair representing the contour of the face of the person, based on the above-mentioned determination result of the line symmetry similarity and the presence or absence of the line corresponding to the boundary with the head. The accuracy is determined, and it is determined whether or not there is a line pair that can be determined as a line pair corresponding to the contour of the face among the line pairs extracted in step 196. If the determination in step 206 is negative, the process proceeds to step 208. If the determination in step 206 is affirmative, the line pair that is determined to have the highest accuracy as the line pair representing the outline of the person's face is regarded as the line corresponding to both sides of the person's face. In the next step 207, a weighting coefficient with a positive sign is given to the face candidate area extracted in step 190.

This weighting coefficient increases based on the line pair representing the contour of the person's face, which is determined to have the highest accuracy, and the higher the line symmetry similarity of the line pair, the higher the value. The value can be set to be high when a line corresponding to the boundary with the head is extracted for the line pair. When the process of step 207 is executed, the process proceeds to step 209. On the other hand, step 198
Alternatively, if the determination in step 206 is negative, it means that a line pair with high accuracy, which is a line pair representing the contour of the human face, is not detected in the face candidate area extracted in step 190. In step 208, a weighting coefficient having a negative sign is given to the face candidate area, and the process proceeds to step 209.

In step 209, it is determined whether the above processing has been performed for all face candidate areas. If a plurality of face candidate areas are set by the face candidate area setting processing by the head extraction described above, the determination in step 209 is negative, and the determination in step 190 is affirmative.
By repeating 0 to 209, a weighting coefficient having a positive or negative sign is given to all the face candidate areas depending on the presence or absence of detection of a line pair representing a contour of a person's face with high accuracy. This weighting factor corresponds to the matching (that is, the accuracy) of the face candidate region as a region corresponding to the human face. If the determination in step 209 is affirmative, the face contour matching determination processing ends, and the process proceeds to step 110 in the flowchart of FIG.

In step 110, a matching determination process based on the internal face structure is performed. This processing searches for a shape pattern representing an eye pair present inside the human face as a shape pattern (second shape pattern) peculiar to the internal structure of the face as the second part of the person, Is to determine a face candidate region with high accuracy that is a region corresponding to the face of
This will be described with reference to the flowchart of FIG. Step two
At 10, one of the face candidate areas is taken out. Next step 2
In 12, the search range of the shape pattern representing the eye pair is set. Since the eye pair exists inside the face, the search range is set so as to match the extracted face candidate area. In step 214, out of a large number of black regions extracted by binarization, a black region that can be a candidate for an eye part existing in the search range, that is, as shown in FIG. And ellipse-shaped black regions having a dimensional ratio in the width direction (short axis direction) (long axis / short axis ratio) within a predetermined range are extracted. In step 216, out of the black areas extracted in step 214, all black area pairs that can be candidates for eye part pairs are extracted. Specifically, the angle in the major axis direction is obtained based on the first moment of inertia of each of the black regions extracted in step 214, and all pairs of black regions in which the difference in the major axis direction angle is within a predetermined range are extracted. .

In the next step 218, it is determined whether or not there is a pair of black areas which can be a candidate eye pair by the above extraction processing. When the determination in step 218 is affirmative, the process proceeds to step 220, and for all the pairs of black areas extracted in step 216, as shown in FIG. 12B, a line connecting the centers of gravity of both black areas. The axis of line symmetry perpendicular to is set, the expansion area for matching is set around each black region, and the line symmetry similarity of the pair of black regions is determined in step 222. In the next step 224, the probability that each black region pair is a black region pair representing an eye pair is determined based on the determination result of the line symmetry similarity of each black region pair, and the black region extracted in step 216 is determined. It is determined whether or not there is a black area pair that can be determined as a black area pair corresponding to the eye pair among the pairs.

If the determination in step 224 is negative, the process proceeds to step 227. When the determination in step 224 is affirmative, the pair of black areas that is determined to have the highest probability of being the pair of black areas that represent the pair of eyes of the person is the pair of black areas that corresponds to the pair of the eyes of the person. Then, in the next step 226, a weight coefficient having a positive sign is given to the face candidate area extracted in step 210. This weighting coefficient is based on the black area pair that is determined to have the highest probability of being the black area pair that represents the eye pair of a person, and the value increases as the line symmetry similarity of the black area pair increases. Can be set to. When the processing of step 226 is executed, step 2
Move to 28.

On the other hand, step 218 or step 224
If the determination is negative, it means that a black region pair having a high probability of being a black region pair representing a person's eye pair is not detected in the face candidate region extracted in step 210. Then, a weighting coefficient having a negative sign is added to the face candidate area, and the process proceeds to step 228.

In step 228, it is determined whether or not the above processing has been performed for all face candidate areas. If a plurality of face candidate areas are set, the determination in step 228 is denied, and the determination in step 228 is affirmative until step 21.
0 to 228 are repeated, and a weighting coefficient having a positive or negative sign is given to all the face candidate areas depending on the presence or absence of detection of a highly accurate black area pair that is a pair of black areas representing a human eye pair. To do. This weighting factor corresponds to the matching (that is, the accuracy) of each face candidate region as a region corresponding to a human face. If the determination in step 228 is affirmative, the consistency determination process based on the internal face structure is terminated, and the process proceeds to step 112 in the flowchart of FIG.

In step 112, a matching determination process based on the body contour is performed. This processing is performed as a shape pattern (second shape pattern) peculiar to the body as the second part of the person, that is, the outline of the body of the person (more specifically, the neck to shoulder, the shoulder to arm portion, or the lower part of the body of the person). (Continuous contour over)
This is for searching a shape pattern that represents the face pattern and determining a face candidate region having a high degree of accuracy, which is a region corresponding to the face of a person, and will be described below with reference to the flowchart of FIG. 7.

In step 230, one of the face candidate areas is extracted, and in step 232, the size and orientation of the extracted face candidate area and the positional relationship between the face of the person and the second portion (outline of the body) are determined. , The search range of the shape pattern representing the body contour is set. As an example, when the extracted face candidate area is in a range surrounded by a broken line in FIG. 13A, the contour of the body (the person's neck to shoulder, shoulder to arm portion or lower body portion) with respect to this face candidate area. (Continuous contour over)
It is estimated that there is a shape pattern that represents
A hatched range in (A) is set as a search range.

In the next step 234, the step 174 of the flowchart of FIG. 4 is performed for the area existing in the search range among the many areas extracted by the binarization.
Similarly to 176, the contour of each area is traced, the curvature of the contour is calculated, and the concave portion and the convex portion (line) are extracted.
Thus, for example, with respect to the original image shown in FIG. 8A, the lines existing in the search range shown in FIG.
Many lines will be extracted.

In step 236, all the sets of lines that can be candidates for one side of the body contour are extracted. Specifically, the distance between the end points of both lines is short, and the angle at which both lines intersect is searched for a pair of lines within a predetermined range, and a set of lines that match the above conditions becomes one side candidate of the body contour. All are extracted as a set of obtained lines. For example, FIG. 13 (B)
When the line as shown in FIG.
As shown in (C), a pair of lines whose intersecting angles are within a predetermined range (θ1 and θ2) and a pair of lines are extracted. For the extracted line set, both lines are extended and connected to generate a one-sided candidate of the body contour.

In step 238, it is determined whether or not a set of lines that can be one side candidate of the body contour has been extracted. If the determination in step 238 is affirmative, step 24
The process proceeds to 0, and for the one-sided candidate body contours generated above, a pair in which the recesses of the one-sided candidate body contours face each other is searched for, and the pair of one-sided candidate that meets the above conditions is set as the body contour candidate. All are extracted as a pair of one-sided candidate body contours. In step 242, it is determined whether or not there is a pair of one-sided candidates that can be candidates for the body contour.

When the determination in step 242 is affirmative, the process proceeds to step 244, and FIG. 13C shows the pairs of one-sided candidates (body contour candidates) of all the body contours extracted in step 240. Set each line symmetry axis as follows,
In the next step 246, the line symmetry similarity of the body contour candidate is determined. Further, a line (for example, the line in FIG. 13B) estimated to correspond to the contour of the person's neck and shoulders forms an angle with the line symmetry axis (for example, FIG. 13).
It is determined whether or not ψ1) in the case of the line (B) is within a predetermined range.

In step 248, the probability that each body contour candidate is a line pair representing the contour of the body of the person is determined based on the determination result of the line symmetry similarity, and among the body contour candidates extracted in step 240. It is determined whether there is a body contour candidate that can be determined as a line pair corresponding to the body contour. When the determination in step 248 is affirmative, the body contour candidate that is the line pair representing the contour of the body of the person and is determined to have the highest accuracy is regarded as the line corresponding to the contour of the body of the person, and the next In step 250, the weighting coefficient having a positive sign is given to the face candidate area extracted in step 230. This weighting factor is based on the line pair (body contour candidate) determined to have the highest accuracy, which is the line pair representing the body contour of the person, and the value increases as the line symmetry similarity of the body contour candidate increases. It can be set higher. When the process of step 250 is executed, the process proceeds to step 254.

On the other hand, step 238 or step 24
2 or if the determination in step 248 is negative,
This is a case where a highly reliable line pair that represents the contour of the body of the person is not detected in the face candidate area extracted in step 230. Therefore, in step 252, the face candidate area has a negative sign. Weighting coefficient is added, and the process proceeds to step 254.

In step 254, it is determined whether or not the above process has been performed for all face candidate areas. If a plurality of face candidate areas are set, the determination in step 254 is denied, and step 23 is performed until the determination in step 254 is denied.
By repeating 0 to 254, a weighting coefficient having a positive or negative sign is given to all face candidate areas depending on whether or not a highly reliable line pair that is a line pair representing the contour of the body of the person is detected. This weighting factor corresponds to the matching (that is, the accuracy) of each face candidate region as a region corresponding to a human face. If the determination in step 254 is affirmative, the matching determination process based on the body contour is ended, and the process proceeds to step 118 in the flowchart of FIG.

In step 118, the face area is comprehensively judged. That is, for each face candidate area set in step 106, the weighting factors assigned in steps 108 to 122 are integrated (or a plurality of weighting factors assigned in each process are divided into positive and negative signs). After multiplication, they are added together to calculate the overall weighting factor of each face candidate region, and the region with the highest overall weighting factor is determined as the region with the highest probability of being the face region of the person. To do.

In the next step 120, step 102
It is determined whether or not the processes of to 118 are executed a predetermined number of times. If the determination in step 120 is negative, step 12
In step 2, various parameters (threshold used for binarization, length of vector for calculating contour curvature (see FIG. 8C), etc.) are changed, and the processes after step 102 are executed again. . In step 122, step 10
Various control parameters may be changed in order to change the coarse density of image division in 2. Step 1
When the processing of 02 to 118 is executed a predetermined number of times, step 120
Is affirmed, the process proceeds to step 124.

In step 124, based on the face areas determined in a predetermined number of times and the weighting factors respectively given to them, the area having the highest accuracy, which is the area corresponding to the human face area, is finally determined. A final determination is made in the same manner as in step 118 above. Then, in the next step 126, as the determination result, the R, G, and B photometric data of the finally determined region is output to the exposure amount calculation unit 44, and the face region extraction processing is ended.

In the face area extraction processing described above, the face candidate area estimated to correspond to the face of a person is not judged based on the density or color of the area. Therefore, even if the color balance of the area changes according to the film type of the negative film 12, the type of light source, the shooting condition such as whether or not there is a backlight, the result of the face area extraction process may change due to this influence. There is no. The face area extraction process described above can be applied even if the original image is a monochrome image. Further, of the face area extraction processing described above, the processing of detecting a shape pattern peculiar to each part of a person (a part of the processing of steps 106 to 112 in the flowchart of FIG. 2) is basically a concave portion and a convex portion, respectively. And extraction of line symmetry similarity. Therefore, if software and hardware are shared by utilizing this, it is possible to realize simplification of software and simplification of hardware configuration.

Further, in the above, the processing target range is limited by performing the non-human region exclusion process, and after the face candidate region is set based on the shape pattern representing the contour of the head of the person, the face candidate is removed. Since the shape pattern representing the contour of the face, the shape pattern representing the internal structure of the face, and the shape pattern representing the contour of the body existing within the search range set according to the area are searched, the entire image is extracted for each shape pattern. The processing time can be shortened as compared with the case where a search is performed as a target, and a highly accurate region corresponding to a human face can be extracted at high speed.

On the other hand, in the exposure amount calculation section 44, the R, G and B photometric data extracted by the face area extraction section 40 as described above and the screen average of one frame calculated by the average density calculation section 42 are calculated. Using the density D _i (i = one of R, G, and B) and the appropriate exposure amount E _i are calculated according to the following equation, and output to the driver 46. Based on the proper exposure amount E _i , the driver 46 sets the color correction filter 18 at a position corresponding to the proper exposure amount.
To move.

LogE _i = LM _i · CS _i · (DN _i −D _i ) + PB _i + LB _i + MB _i + NB _i + K ₁ + K ₂ (1) However, the meanings of the respective symbols are as follows.

LM: Magnification slope coefficient. It is set in advance according to the enlargement magnification determined by the type of negative film and the print size.

CS: Color slope coefficient. It is prepared for each type of negative film, and there are one for underexposure and one for overexposure. It is determined whether the average density of the image frame to be printed is under or over the standard negative density value, and either under exposure or over exposure is selected.

DN: Standard negative density value. D: Density value of image frame to be printed.

PB: Correction balance value for standard color paper. It is determined according to the type of color paper.

LB: Correction balance value for standard printing lens. It is determined according to the type of lens used for printing.

MB: Correction value (master balance value) for fluctuations in light source light quantity and paper development performance.

NB: Negative balance (color balance) value determined by the characteristics of the negative film.

K ₂ : Color correction amount. K ₁ : A density correction amount represented by the following formula.

[0094]

[Equation 1]

Here, Ka and Kb are constants, and FD is the face area average density. Further, the density correction amount K ₁ in the above equation (8) may be used as the correction value obtained by the film inspection apparatus, and the color correction amount K ₂ may be expressed using the face area average density as follows.

[0096]

[Equation 2]

However, Kc is a constant. Further, the density correction amount K ₁ and the color correction amount K ₂ in the above formula (1) are used as the correction amounts obtained by the film inspection device, and the average density D _i of the print frame in the formula (1) is the average density FD of the face area. _i
Alternatively, the exposure amount may be calculated.

Although the negative film 12 is described as an example of the recording medium for the original image in the above, an image recorded on another film such as a positive film or various recording media such as paper may be used as the original image. It is possible. Further, although color paper has been described as an example of the copying material, other materials such as paper may be applied.

Further, in the above, in the face candidate area setting process by head extraction, as the shape pattern peculiar to the head of the person, the convex portion corresponding to the crown of the person and the boundary between the head and the face of the person are corresponded. Although the concave portion is used, the invention is not limited to this. For example, in order to extract a head without hair or a head with low hair density, area division by a method other than binarization or detection of an edge of an image is performed to form a convex shape pattern unique to the head. The head may be extracted using only the part.

Further, although binarization has been described above as an example of a method of dividing an image into a plurality of areas, it is not limited to this and is described in Japanese Patent Laid-Open No. 4-346332. You may divide using a histogram like this.
In addition, the density value of each pixel of the image is different from each other in a plurality of different directions (for example, as shown in FIG. 14A, the horizontal direction (0 °), the vertical direction (90 °), and the horizontal direction of the image). ±
The edge of the image is detected by differentiating along the direction (inclined by 45 °), and instead of the line described in the above embodiment, the center of the edge obtained by performing a known thinning process on this edge. You may make it process using a line.

The differentiation along each of the above directions is shown in FIG.
This can be realized by using each of the four types of differential filters shown in (B), and the detection of the edge of the image is such that the pixel of interest is an edge when the magnitude of the differential filter output is larger than a predetermined threshold value. Can be considered Further, the edge extraction processing may be repeatedly performed while changing the threshold value for edge determination stepwise.

In the above description, the case where the eye pair is used as the shape pattern representing the internal structure of the face has been described as an example. However, the shape pattern is not limited to this, and a pair of nose, mouth, and eyebrows may be used. You may make it set a face candidate area | region comprehensively combining these.

Further, in the above description, as the shape pattern representing the body contour, an example of extracting a pair of lines corresponding to the contours continuous from the neck to the shoulder of the person, from the shoulder to the arm portion or the lower part of the body has been described. Instead, the pair of lines corresponding to the contour of the person's neck and shoulder and the pair of lines corresponding to the contour of the person's shoulder and arm may be detected separately.

In the above description, the shape pattern representing the outline of the person's head is used as the first shape pattern, and the shape pattern representing the outline of the person's face is used as the second shape pattern.
Although the example using the shape pattern representing the internal structure of the person's face and the shape pattern representing the body structure of the person has been described, the present invention is not limited to this. For example, a shape pattern representing the contour of the head of the person. A plurality of shape patterns may be used as the first shape pattern, such as a shape pattern representing the internal structure of the face.

The embodiments of the present invention have been described above.
In addition to the technical matters described in the scope of claims, the above-described embodiments include embodiments of the technical matters described below.

(1) The search for the shape pattern representing the contour of the head of the person as the first shape pattern or the second shape pattern is performed by the contour of the area divided by the invention according to claim 3, or According to the invention described in 4, the concave portion and the convex portion are detected from the center line of the edge, and the feature amount (circularity, position, size,
The exposure amount determining method is performed by determining whether or not the concave portion or the convex portion corresponds to the contour of the head of the person based on the (direction).

(2) The search for the shape pattern representing the contour of the person's face as the first shape pattern or the second shape pattern is performed by the contour of the area divided by the invention according to claim 3 or claim 4. The relationship between the feature amounts (circularity, position, size, direction) of each of the two or more recesses or protrusions detected by detecting the recesses and protrusions from the center line of the edge obtained by the described invention Based on the degree of line symmetry, it is determined whether or not the two or more concave portions or convex portions correspond to the contour of the human face based on the degree of line symmetry.

(3) The search for the shape pattern representing the eye pair among the shape patterns representing the human face structure as the first shape pattern or the second shape pattern is divided by the invention according to claim 3. Area, or an elliptical area of the area divided by the center line of the edge obtained by the invention according to claim 4, and the feature amount (shape, density) of each of the extracted elliptic areas is extracted. Detects a pair of elliptic areas having a similarity of a predetermined value or more by a matching process using, and the pair of elliptic areas corresponds to the eye pair of a human face according to the degree of line symmetry of the detected pair of elliptic areas. A method for determining an exposure amount, which is performed by determining whether or not there is an exposure amount.

(4) The search for the shape pattern representing the contour of the body of the person as the first shape pattern or the second shape pattern is performed by the contour of the area divided by the invention according to claim 3 or The concave portion and the convex portion are detected from the center line of the edge obtained by the invention described above, and the characteristic amounts (circularity, position, size,
Direction), a plurality of sets of concave portions or convex portions corresponding to the contour from the neck to the shoulder of the person and the contour from the shoulder to the arm portion or the lower part of the body are detected, and further the line symmetry between the pairs is detected. A method of determining an exposure amount, which is performed by determining whether or not it corresponds to the contour of the body of a person based on the degree.

(5) As a region with low accuracy corresponding to a person in the original image, a region divided by the invention according to claim 3 or a center line of the edge obtained by the invention according to claim 4 is divided. Of the regions, the ratio of the linear portion included in the contour of the region is a predetermined value or more, or the line symmetry degree is a predetermined value or less, or the number of irregularities is a predetermined value or more,
Or, the area where the contact rate with the outer edge of the image is a predetermined value or more, or the area where the contrast of the internal density is a predetermined value or less, or the internal density is changing in a predetermined pattern, or a predetermined change pattern is repeated. A method for determining an exposure amount, which comprises excluding an area.

[0111]

As described above, according to the first aspect of the invention, the first shape pattern unique to the first part of the person existing in the original image is searched for and detected. After setting an area estimated to correspond to the person's face according to the size, orientation, and positional relationship between the person's face and the first portion, the characteristic of at least one second portion of the person is set. By searching for whether the second shape pattern exists within the search range set according to the size and orientation of the set area in the original image and the positional relationship between the face of the person and the second portion. since the probability is a region corresponding to the face of a person in the original image was high area judgment be so that, influenced by the color of the region and its peripheral region corresponding to the face of the person in the original image without, it can extract only an area corresponding to a person's face , Has an excellent effect that.

According to a third aspect of the present invention, the original image is divided into a plurality of regions composed of a plurality of pixels having the same or similar density or color, and the shape is formed by using the contour of each of the plurality of regions. Since the pattern is searched, there is an effect that the shape pattern peculiar to each part of the person can be efficiently detected.

According to the fourth aspect of the present invention, the edge existing in the original image is detected, the center line of the detected edge is obtained, and the shape pattern is searched using the obtained center line. This has the effect that the shape pattern can be efficiently detected.

According to the fifth aspect of the present invention, prior to the search for the first shape pattern, an area with low accuracy corresponding to a person in the original image is excluded from the search object for the first shape pattern. The time required for the shape pattern search processing is shortened, and the processing can be performed at a higher speed. The invention according to claim 7 is from claim 1 to claim
Since the exposure amount to the copy material is determined based on at least one of the color and the density of the highly accurate region which is the region corresponding to the face of the person in the original image, which is determined by any one of the face region extraction methods of 6. The exposure amount can be determined so that the area corresponding to the person's face in the original image can be properly printed without being affected by the colors of the area corresponding to the person's face and the surrounding areas. Has excellent effect.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a photographic printing apparatus according to this embodiment.

FIG. 2 is a flowchart illustrating a main routine of face area extraction processing executed by a face area extraction unit.

FIG. 3 is a flowchart illustrating a non-person area determination process.

FIG. 4 is a flowchart illustrating face candidate area setting processing by head extraction.

FIG. 5 is a flowchart illustrating a matching determination process based on a face contour.

FIG. 6 is a flowchart illustrating a matching determination process based on an internal face structure.

FIG. 7 is a flowchart illustrating a matching determination process based on a body contour.

8A and 8B show details of a face candidate area setting process by head extraction, where FIG. 8A is an original image, FIG. 8B is a contour trace of a black area extracted by binarization, and FIG. 8C is a contour curvature. Calculation,
(D) is an image diagram for explaining extraction of a concave portion and a convex portion, respectively.

FIG. 9 shows details of the consistency determination process by head extraction.
(A) is an image diagram for explaining the calculation of the feature amount of the concave portion and the convex portion, (B) is a determination of the head, and (C) is an image diagram for explaining the setting of the face candidate region.

FIG. 10 shows details of the matching determination process based on the face contour.
(A) sets a face contour search range, (B) extracts concave and convex portions, (C) extracts line pairs that can be face contour candidates, and (D) determines line symmetry similarity. It is an image figure for each explaining.

FIG. 11 shows details of the consistency determination process based on the face contour.
(A) is an image diagram for explaining the determination of line symmetry similarity, and (B) is an image diagram for explaining a search for a boundary with the head.

FIG. 12 shows details of the consistency determination process based on the face structure.
(A) is an image diagram for explaining extraction of a black region that can be a candidate for an eye part, and (B) is a diagram for explaining determination of a line symmetry similarity of a pair of black regions.

13A and 13B show details of a body contour matching determination process, (A) setting of a body contour search range, (B) extraction of concave and convex portions, and (C) determination of line symmetry similarity. It is an image figure for each explaining.

14A is a conceptual diagram showing an example of a differential direction in edge detection, and FIG. 14B is a conceptual diagram showing an example of a differential filter for performing differentiation in each direction.

[Explanation of symbols]

10 Photo printing equipment 12 Negative film 18 color correction filter 28 color paper 32 CCD image sensor 40 Face Area Extraction Unit 44 Exposure amount calculation unit

Claims (7)

(57) [Claims] 1. A first shape pattern unique to a first part of a person existing in an original image is searched for, and the detected size and orientation of the first shape pattern and the face of the person and the first shape pattern are detected. After setting a region that is estimated to correspond to the face of the person according to the positional relationship with the portion of, the second shape pattern peculiar to at least one second portion different from the first portion of the person is set. , The size of the set area in the original image, the orientation and by searching for a position within the search range set according to the positional relationship between the face of the person and the second portion, A face area extraction method for determining an area having a high degree of accuracy, which is an area corresponding to a human face. 2. The first shape pattern and the second shape pattern are a shape pattern representing a contour of a human head, a shape pattern representing a contour of a human face, and a shape pattern representing an internal structure of a human face. The face area extracting method according to claim 1, further comprising at least one of a shape pattern representing the contour of the body of the person. 3. The search for the first shape pattern and the second shape pattern is performed by dividing an original image into a plurality of regions each composed of a plurality of pixels having the same or similar density or color,
The face area extracting method according to claim 1, wherein the contour area is extracted using contours of the plurality of areas. 4. The search for the first shape pattern and the second shape pattern is performed by detecting an edge existing in the original image, obtaining a centerline of the detected edge, and using the obtained centerline. The face area extracting method according to claim 1, wherein 5. The region of low accuracy corresponding to a person in the original image is excluded from the search target of the first shape pattern prior to the search of the first shape pattern. Face Area Extraction Method. 6. The first shape pattern or the second shape pattern
A shape pattern that represents the facial structure of a person as a shape pattern
The search for the shape pattern that represents the eye pair among the search results is described in claim 3.
The divided areas according to the invention described in claim 4, or claim 4.
Divided by the center line of the edge found by the invention
An elliptical area is extracted from the area, and the extracted multiple
By matching processing using each feature of the elliptical area
A pair of elliptical areas with a similarity of a predetermined value or more is detected and detected.
Depending on the degree of line symmetry of the pair of elliptic regions, the pair of elliptic regions
Can correspond to the eye pairs of the human face.
The method for extracting a face area according to claim 1, wherein
Law. 7. The method according to any one of claims 1 to 6.
The face area extraction method of
Area that has a high degree of accuracy, and based on at least one of the color and density of the area
An exposure amount determining method, characterized in that an exposure amount to a copying material is determined.