JP2008287704A - Face image detection device, face image detection method and photographing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a face image detection device, capable of accurately detecting a face image position without increasing the detection time of the face image position and the power consumption needed for detecting the face image position. <P>SOLUTION: The face image detection device for detecting a face image position using input image data comprises a reduced and rotated image data formation part 104 which reduces and rotates the input image data by 360×i/n °(n is an integer of 2 or more, i=0 to n-1) to form n-pieces of reduced and rotated image data at every 360/n°; and a face detection part 106 which detects the face image position for the n-pieces of reduced rotated image data. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for improving the speed and accuracy of a face detection function used when, for example, a human camera takes a human image, and in particular, a face image provided in an image recording apparatus such as a digital camera or a personal computer. The present invention relates to a detection device.

  Human face position detection technology has been used not only in surveillance systems but also in consumer digital still cameras and the like. There is a remarkable increase in technology and spread of high-speed and high-quality digital cameras. For example, in an autofocus function for a person (especially a face) in a digital camera, an automatic exposure correction function, and personal authentication based on face recognition, the performance of the digital still camera has been improved and a face position detection technique has been put into practical use. Yes. In addition, the face position detection technique is useful as a technique for extracting a person from a moving image or a still image in image search, indexing, and image album organization in digital video, and its application development is wide.

As a face position detection technique capable of improving the position detection accuracy regardless of the face orientation, there is a conventional technique such as Patent Document 1. In Patent Document 1, when detecting the position of one face, the front face image parameter is not used, the front face image is divided into a left half and a right half, and two sets of parameters relating to the left and right face images are set. Use to detect the position of the face image. As a result, the position detection rate is improved even for a sideways face other than the front face.
JP 2006-72770 A

  However, in Patent Document 1, since the position of the face image is detected using two sets of parameters as described above, in order to detect the position of the face image from one input image, a set of parameters is used. It is necessary to execute a reference comparison process between a set of parameters and an input image every time. That is, it is necessary to execute the reference comparison process twice. For this reason, there are problems such as the detection time of the face image position and the increase in power consumption required for detecting the face image position.

  Further, in Patent Document 1, the position of a face image can be detected with high accuracy when a person turns to the left or right. However, when the face image rotates with the optical axis of the digital camera as the central axis, it is possible to accurately There is also a problem that the position of the face image cannot be detected.

  The present invention has been made to solve the above-described problems, and can accurately detect the position of a face image without increasing the detection time of the face image and the power consumption required for detecting the face image position. An object of the present invention is to provide a face image detection device capable of performing the above.

  A face image detection apparatus according to the present invention is a face image detection apparatus that detects the position of a face image using input image data, and reduces the input image data to 360 × i / n (n is 2 or more). Integer: i = 0 to n-1) Reduced rotated image data creating means for creating n reduced rotated image data every 360 / n degrees rotated, and the n reduced rotated image data as targets. And face detection means for detecting the position of the face image.

  According to this configuration, the position of the face image is detected from a plurality of reduced rotated image data obtained by reducing the input image data and rotating the input image data at a plurality of angles. For this reason, the position of the face image can be detected with high accuracy. The reduced rotation image data is obtained by reducing the input image data. For this reason, the position of the face image can be detected at high speed, and the power consumption can be reduced.

  Preferably, the face detecting means uses a technique capable of detecting the position of a face image having an inclination within a range of at least 360 / n degrees, and the position of the face image with respect to the n pieces of reduced rotated image data. Is detected.

  Since the face detection means can detect the position of the face image having an inclination within a range of 360 / n degrees, the face detection means detects the position of the face image from n pieces of reduced rotation image data, and thereby the inclination is 360 degrees. In contrast, the position of the face image can be detected.

  More preferably, the above-described face image detection device further specifies reduced rotation image data specifying the reduced rotation image data in which the position of the face image is detected by the face detection unit among the n reduced rotation image data. Based on the rotation angle θ of the reduced rotated image data specified by the reduced rotated image data specifying means with respect to the input image data, out of n angle setting values every 360 / n degrees (− angle setting value selecting means for selecting one angle setting value used for position detection of a face image having an inclination of θ), and the face detecting means is further selected by the angle setting value selecting means. The position of the face image is detected for the input image data using one angle setting value.

  Since the face image position detection process is performed on the input image data using any of the n angle setting values, highly accurate face image position detection can be realized. For example, when the ratio of the face to the image is small, the position of the face image can be detected from the input image data after determining the rough inclination of the face image from a plurality of reduced rotation image data. Thereby, the face image position detection process can be performed using the optimum angle setting value, and the position of the face image is detected with high accuracy.

  More preferably, when there are a plurality of the reduced rotated image data specified by the reduced rotated image data specifying means, the angle setting value selecting means further corresponds to the input image data for each of the plurality of reduced rotated image data. Based on a plurality of rotation angles θ, a plurality of angle setting values used for position detection of a face image having an inclination of (−θ) are selected from n angle setting values every 360 / n degrees, The face detection unit may further include, when there are a plurality of the reduced rotated image data specified by the reduced rotated image data specifying unit, from among the plurality of angle setting values selected by the angle setting value selecting unit. One angle set value is taken out in order, and the position of the face image is detected for the input image data using the taken out one angle set value.

  If there is a face image having an inclination angle substantially the same as the allowable angle at which the position of the face image can be detected by the face detection means, two reduced rotation image data may be specified. In such a case, The position of the face image can be detected based on one angle setting value while alternately switching the two angle setting values corresponding to the two specified reduced rotation image data. Thereby, the position of the face image can be detected with high accuracy.

  More preferably, the face detecting means detects the position of the face image again for the n pieces of reduced rotated image data when the reduced rotated image data is not specified by the reduced rotated image data specifying means. It is characterized by that.

  When the specified reduced rotation image data is lost, it is determined whether the inclination of the face image has changed or there is no face image in the input image data, and n reduced rotation image data are again obtained. The position of the face image is detected for. Thus, for example, even when shooting a movie or still image, even if the position and inclination of the face image changes due to the subject moving or the photographer moving the hand, Following the change, the position of the face image can be detected at high speed and with high accuracy.

  More preferably, the face detection means uses the predetermined angle setting value used for detecting the position of a face image having a predetermined angle of inclination as a target for the n pieces of reduced rotated image data. The face image detection device detects a position of the image, and the face image detection device further specifies a reduced rotation image data that specifies the reduced rotation image data from which the position of the face image is detected by the face detection unit among the n reduced rotation image data. The face detection means further includes a predetermined angle setting value when there are a plurality of the reduced rotated image data specified by the reduced rotated image data specifying means. The position of the face image is detected for a plurality of the reduced rotated image data specified by the reduced rotated image data specifying means.

  If there is a face image having an inclination angle substantially the same as the allowable angle at which the position of the face image can be detected by the face detection means, two reduced rotation image data may be specified. In such a case, The position of the face image can be detected from the two specified reduced rotation image data.

  More preferably, the face detection means uses the predetermined angle setting value used for detecting the position of a face image having a predetermined angle of inclination as a target for the n pieces of reduced rotated image data. It is characterized by detecting the position of an image.

  Since the position of the face image is detected using one angle setting value, the position of the face image can be detected at high speed, and power consumption can be reduced.

  More preferably, the reduced rotation image data creation unit further creates one face detection image including the n pieces of reduced rotation image data, and the face detection unit includes the one face detection image. The position of the face image is detected for the target.

  Face image position detection processing can be performed on one piece of face detection image data including face images with a plurality of inclinations at once. For this reason, the position of the face image can be detected at high speed. Accordingly, the power consumption of the face image detection device can be reduced.

  Note that the present invention can be realized not only as a face image detection apparatus including such characteristic means but also as a face image detection method using the characteristic means included in the face image detection apparatus as a step. Alternatively, it can be realized as a program that causes a computer to execute characteristic steps included in the face image detection method. Needless to say, such a program can be distributed via a recording medium such as a CD-ROM (Compact Disc-Read Only Memory) or a communication network such as the Internet.

  Further, it can also be realized as a photographing device equipped with a face image detection device.

  According to the present invention, it is possible to improve position detection speed, reduce power consumption, and improve position detection accuracy even in face image position detection when the face of a person in the image is inclined. For example, even with a composition in which a baby is held, high-speed and high-accuracy and stable face image position detection can be performed. In addition, it is possible to provide a digital camera capable of operating an autofocus function for a person or face and an automatic exposure correction function with high accuracy.

  Hereinafter, a face image detection apparatus according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a face image detection apparatus according to an embodiment of the present invention.
The face image detection apparatus 100 includes an A / D (Analog to Digital) converter 101, a face image detection unit 111, and a CPU (Central Processing Unit) 102.

  The A / D converter 101 is a processing unit that converts an analog image data signal of an input image into a digital image data signal. Note that a DSP (Digital Signal Processor) provided in the A / D converter 101 converts a digital image data signal into a standard signal REC601 image data signal, REC656 image data signal, or a general-purpose input format. You may output, after converting into the combined image data signal.

  The face image detection unit 111 is a processing unit that detects the position of the face image included in the input image data from the image data signal (input image data) output from the A / D converter 101, and the luminance signal extraction unit 103. A reduced rotation image data creation unit 104, a memory 105, a face detection unit 106, and a CPU interface unit 107.

  The luminance signal extraction unit 103 is a processing unit that extracts luminance signals from input image data and creates luminance signal data.

  The reduced rotation image data creation unit 104 performs data filtering processing and reduction resizing processing on the luminance signal data extracted by the luminance signal extraction unit 103, and data of a prescribed size used for face image position detection (hereinafter referred to as “face”). This is a processing section that writes “detection image data”) into the memory 105.

  The memory 105 is a storage device for storing the luminance signal data of the input image data extracted by the luminance signal extraction unit 103 and the face detection image data generated by the reduced rotation image data generation unit 104.

  The face detection unit 106 is a processing unit that accesses the luminance signal data or the face detection image data stored in the memory 105 and detects the position, size, and inclination of the face image. The face detection unit 106 stores four types of face detection parameters that are optimal for detecting the face image position and the like in a state where there is no left-right tilt. Based on one of the types of face detection parameters, Detection of the face image position and the like is performed. Since the processing of the face detection unit 106 can use a known technique and is not the main point of the present application, detailed description thereof will not be repeated here. The face detection parameter is a parameter obtained by quantifying the luminance distribution in a general face image. The four types of face detection parameters stored in the face detection unit 106 indicate face detection parameters when the angle setting values indicating the inclination of the face image are 0 degrees, 90 degrees, 180 degrees, and 270 degrees. . For example, a face detection parameter having an angle setting value of 90 degrees is a face detection parameter that is optimal for detecting the position of a face image tilted by 90 degrees.

  The CPU interface unit 107 is a processing unit that receives a command from the CPU 102 and controls system operations of the luminance signal extraction unit 103, the reduced rotation image data creation unit 104, and the face detection unit 106.

  The CPU 102 is a processing unit that controls the system operation of the face image detection unit 111 via the CPU interface unit 107, and is functionally reduced (as a processing unit realized by executing a control program on the CPU 102). An image data specifying unit 108, an angle setting value selection unit 109, and a face image position determination unit 110 are included.

  The reduced rotated image data specifying unit 108 is a processing unit that specifies data in which the position of the face image is detected by the face detecting unit 106 among the data resized by the reduced rotated image data creating unit 104.

  The angle setting value selection unit 109 is a processing unit that sets a face detection parameter including an angle setting value used for face image position detection in the face detection unit 106.

  The face image position determination unit 110 inputs the input image data from the position of the face image detected by the face detection unit 106 and the correspondence relationship between the data specified by the reduced rotation image data specifying unit 108 and the input image data. The position of the face image at is determined.

Next, the operation of the face image detection apparatus 100 will be described.
Every time input image data is obtained, the luminance signal extraction unit 103 extracts a luminance signal used for face image position detection from the input image data, and creates luminance signal data.

  FIG. 2 is a flowchart showing operations of the CPU 102, the reduced rotation image data creation unit 104, the memory 105, the face detection unit 106, and the CPU interface unit 107 of the face image detection apparatus 100.

  In the following description, the processing of S4 to S9, the processing of S10 to S16, and the processing of S10 and S18 to S22 are each performed within one frame period (for example, 1/30 seconds).

  The luminance signal output from the luminance signal extraction unit 103 is luminance signal data of input image data having an arbitrary size. For this reason, the reduced rotation image data creation unit 104 performs luminance processing (face detection image data) on which filtering processing and reduction resizing processing are performed in order to convert the luminance signal data into an arbitrary size specified by the face detection unit 106. And the face detection image data is written to the memory 105 in a prescribed size (S4).

  The face detection unit 106 accesses the image data for face detection stored in the memory 105, and detects the position, size, and inclination of the face image (S6).

  Here, the face detection unit 106 is set with one type of face detection parameter that is optimal for face image position detection in a state where there is no left-right tilt, and face image position detection using this face detection parameter. In the processing, it is assumed that the position of the face image tilted up to 45 degrees on each of the left and right sides can be detected (hereinafter referred to as “tilt allowable angle is 45 degrees”). In this case, in the process of S4, the reduced rotation image data creation unit 104 uses the luminance signal data of the input image data (hereinafter, referred to as “original image data”) as 1 / n (the specified size of the face detection image data). n = 4) Reduced to below to create reduced image data. Further, the reduced rotation image data creation unit 104 has one face obtained by pasting together the four reduced rotation image data obtained by rotating the reduction image data by 360 / n (n = 4), that is, by 90 degrees. Image data for detection is generated. For this purpose, the reduced rotation image data creation unit 104 performs mapping of the four reduced rotation image data within the specified size address of the face detection image data while rotating the reduced image data by 90 degrees, and performs four reduction rotations. Image data is written to the memory 105 at high speed. The amount of data written to the memory 105 is one specified size of face detection image data. For this reason, when writing the four pieces of reduced rotated image data, the data amount is the same as when writing the original image data as it is, and the power increase is slight.

  FIG. 3 is a diagram illustrating an example of original image data. The original image data 201 is image data obtained by imaging a baby as a subject, the tilt angle of the subject's face is 60 degrees, and exceeds the allowable tilt angle of the face image position detection process in the face detection unit 106. .

  FIG. 4 is a diagram illustrating an example of image data for face detection. It is assumed that the prescribed size of the face detection image data 202 is the same as the image size of the original image data 201. The face detection image data 202 is image data in which four pieces of reduced rotation image data 203 to 206 are pasted so as to fit within a specified size. The reduced rotation image data 203 is rotated by 0 degrees with respect to the original image data 201. The reduced rotation image data 204 is rotated 180 degrees with respect to the original image data 201. The reduced rotation image data 205 is rotated 270 degrees clockwise with respect to the original image data 201. The reduced rotation image data 206 is rotated 90 degrees clockwise with respect to the original image data 201.

  In the process of S <b> 6, the face detection unit 106 reads out one piece of face detection image data 202 within a specified size obtained by pasting four types of images from the memory 105, and executes face image position detection processing. By this processing, face image position detection processing for four types of image information can be performed collectively. Even for an image in which the face image position cannot be detected, such as the original image data 201, whose face tilt exceeds 45 degrees, face image position detection processing is performed on four types of reduced images rotated by 90 degrees. The For this reason, the position, size, and inclination of the face image 205a are detected in the reduced rotation image data 205 in which the right / left inclination of the face is within ± 45 degrees.

  Next, the reduced rotation image data specifying unit 108 specifies the reduced rotation image data from which the face image position is detected by the face detection unit 106 among the four types of reduced rotation image data (S8). For example, in the face detection image data 202 shown in FIG. 4, the reduced rotation image data 205 is specified as the reduced rotation image data from which the face image position is detected. Since the arrangement of the four types of reduced rotation image data is predetermined, the reduced rotation image data is specified from the position of the detected face image on the original image data.

  The face image position determination unit 110 detects the position, size, and inclination of the face image on the original image data from the position of the face image in the specified reduced rotation image data (S9). The correspondence between the reduced rotation image data and the original image data is known when the reduced rotation image data creation unit 104 writes the reduced rotation image data in the memory. The reduction rate is also known in advance. Therefore, the position and size of the face image on the original image data can be calculated from these pieces of information. Further, since the rotation angle at the time of creating the reduced rotation image data is also known, the inclination of the face image on the original image data can be calculated from the inclination of the face image in the reduced rotation image data.

  When there is one reduced rotation image data from which the face image position is detected (YES in S10), the angle setting value selection unit 109 detects the face from the reduced rotation image data specified by the reduction rotation image data specification unit 108. By specifying the angle setting value used for the face image position detection process in the unit 106, the face detection parameter including the angle setting value is specified (S12).

  When the face image position is detected in any one of the four types of reduced rotation images, the inclination of the person is determined based on the rotation angle information of each reduced rotation image mapped in the memory 105 by the write control to the memory 105. The composition can be determined. In the example shown in FIGS. 3 and 4, since the position of the face image 205a is detected in the reduced rotation image data 205 obtained by rotating the original image data 201 by 90 degrees counterclockwise, the face image of the subject in the original image data 201 is detected. 201a can be determined to be rotated approximately 90 degrees clockwise within the frame. Therefore, the angle setting value selection unit 109 specifies the angle setting value as 90 degrees, and outputs the angle setting value to the face detection unit 106. As a result, the face detection unit 106 specifies the face detection parameters including the angle setting value used for the next face image position detection.

  The face detection unit 106 accesses the original image data stored in the memory 105 using the face detection parameter including the angle setting value, and detects the position, size, and inclination of the face image (S14). In the above example, the angle setting value is specified as 90 degrees. Therefore, the face detection unit 106 accesses the original image data 201 shown in FIG. 3 using the face detection parameters including the angle setting value 90 degrees, and detects the position, size, and inclination of the face image 201a. To do. Note that the original image data stored in the memory 105 is sequentially updated. Therefore, the face detection unit 106 detects the face image position from the latest original image data.

  While the face image position is detected, the face detection unit 106 detects the position of the face image from the original image data 201 (NO in S16, S14).

  If the face image position is no longer detected (YES in S16), the processes after S4 are repeatedly executed.

  When a face image position is detected from a plurality of reduced rotation image data in the process of S10 (NO in S10), the reduced rotation image data creation unit 104 specifies a plurality of reduction rotations specified by the reduction rotation image data specifying unit 108. The luminance signal data output from the luminance signal extraction unit 103 is subjected to filter processing and reduction / resize processing so that only the image data fits within the specified size of the face detection image data, thereby generating face detection image data. The reduced rotation image data creation unit 104 writes the created face detection image data in the memory 105 (S18). For example, if the position of the face image is detected from the reduced rotation image data 203 and 205 in the face detection image data 202 shown in FIG. 4, the reduced rotation image data creation unit 104 has a configuration as shown in FIG. Then, face detection image data 211 is created by mapping two types of reduced rotation image data 213 and 215 within the prescribed size of the face detection image data.

  The face detection unit 106 accesses the face detection image data stored in the memory 105 in the process of S18, and detects the position, size, and inclination of the face image (S20).

  The face image position determination unit 110 detects the position, size, and inclination of the face image on the original image data from the position of the face image in the specified reduced rotation image data (S21). For example, it is assumed that the position of the face image 205a is detected from the reduced rotation image data 215 of the face detection image data 211, and the inclination of the face 215a is determined to be 330 degrees clockwise (30 degrees counterclockwise). At this time, the reduced rotation image data 215 including the face 215 a is rotated 270 degrees clockwise with respect to the original image data 201. Therefore, the inclination of the face image 201a in the original image data 201 is calculated as 60 (= 330-270) degrees clockwise.

  While the position of the face image is detected, the face detection unit 106 detects the position of the face image from the face detection image data (NO in S22, S18, S20).

  When the position of the face image is no longer detected (YES in S22), the processes after S4 are repeatedly executed.

  As described above, according to the present embodiment, even for a face image having an inclination greater than the allowable inclination of the face, a single type of face detection parameter is used instead of a plurality of types of face detection parameters. The position can be detected at high speed. In other words, after detecting the face image position in the face detection image data, the face image position is detected after limiting the face detection parameter to one, so that the detection speed of the face image position is increased. Can be improved. For this reason, the power consumption of the face image detection apparatus 100 can be reduced.

  Further, since the face image position detection process is performed on the original image data data using any of the four types of face detection parameters every 90 degrees, it is possible to realize highly accurate face image position detection. For example, when the ratio of the face image in the original image data is small, after determining the rough inclination of the face image from the face detection image data pasted with the reduced rotation image data, the original image instead of the face detection image data is determined. The position of the face image can be detected from the image data. Thereby, the position of the face image is detected with high accuracy.

  In addition, during the face image position detection process in which the face detection parameters are limited to one type, if the number of detected face image positions becomes 0, the inclination of the face image has changed, or the input image It is determined whether there is no face image in the data, and the processing after the generation processing of the face detection image data is repeated again. Thus, for example, even when shooting a movie or still image, even if the position and inclination of the face image changes due to the subject moving or the photographer moving the hand, Following the change, the position of the face image can be detected at high speed and with high accuracy.

  The face detection image data is data of a prescribed size used for position detection of the face image. Therefore, data writing to the memory 105 and the operation of the face detection unit 106 can be performed at high speed. Thereby, the power consumption of the face image detection apparatus 100 can be reduced.

  In addition, when there is a face image having an inclination angle substantially the same as the allowable inclination angle in the face detection unit 106, the face image position may be detected from the two reduced rotation image data. In such a case, Even if it exists, the position and inclination of the face image can be detected with high accuracy.

  The face image detection apparatus according to the embodiment of the present invention has been described above, but the present invention is not limited to this embodiment.

  For example, in the operation of the face image detection apparatus 100 shown in FIG. 2, when the face image position is detected from a plurality of reduced rotation image data, the face image position is detected from the face detection image data. The face image position may be detected from the original image data. In this case, the angle setting value selection unit 109 specifies a plurality of angle setting values corresponding to the plurality of reduced rotation image data, thereby specifying a plurality of types of face detection parameters including the angle setting values. . The face detection unit 106 sequentially extracts one type of face detection parameter from the specified types of face detection parameters, and uses the one type of face detection parameter thus extracted as a target for original image data. The position of the face image is detected. Thereby, the position of the face image can be detected with high accuracy.

  In the above-described embodiment, since the allowable tilt angle is 45 degrees, the image data for which the luminance signal data is reduced is rotated by 90 (= 45 × 2) degrees, and the face is detected by four memory writing operations. I was creating image data. However, it goes without saying that the present invention can be applied even if the allowable tilt angle is other than 45 degrees. For example, when the allowable tilt angle is 90 degrees, image data for face detection is generated by two memory writing operations while rotating the image data obtained by reducing the luminance signal data by 180 (= 90 × 2) degrees. You may do it. FIG. 6 is a diagram showing an example of face detection image data 207 created by two memory write operations while rotating the image data obtained by reducing the original image data 201 shown in FIG. 3 by 180 degrees. The face detection unit 106 detects the face image position for the face detection image data 207. Thereby, the face image position is detected from one of the two reduced rotation image data.

  The face image detection apparatus 100 according to the present embodiment can be mounted on a digital camera such as a general digital still camera or digital movie camera.

  FIG. 7 is a diagram illustrating an example of the appearance of a digital still camera equipped with the face image detection device 100. FIG. 8 is a diagram showing an example of the configuration of a digital still camera equipped with the face image detection device 100. As shown in FIG. The digital still camera 300 includes a lens 301, an image sensor 302, a CDS / AGC circuit 303, an A / D converter 101, a face image detection device 100, a display device 304, a recording medium 305, and a timing generator 306. Including.

  The light collected by the lens 301 is converted into a pixel signal by the image sensor 302. The CDS / AGC circuit 303 performs correlated double sampling on the pixel signal output from the image sensor 302, performs automatic gain adjustment processing, and outputs an image data signal. The timing generator 306 causes the image sensor 302 to generate drive pulses at predetermined time intervals. The display device 304 is a display device that displays original image data and the position of a face image detected on the original image data, and is, for example, an LCD (Liquid Crystal Display). The recording medium is a recording device that records original image data.

  By mounting such a face image detection device 100 on a digital camera, the user can take a picture of a person using a highly accurate and stable face image position detection function.

  Further, by using the technique of the present invention, it is possible to accurately detect not only the face image but also the position and line of sight of the eyes included in the face.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

  According to the face image detecting apparatus according to the present invention, the position of the face image of all inclinations can be detected by one face image position detecting operation. Therefore, for example, a face image position in a composition in which a baby is held or an oblique composition of a sleeping face can be detected stably at high speed. Therefore, the present invention can be applied to an imaging apparatus such as a digital camera capable of operating the autofocus function for a person (face) and the automatic exposure correction function at high speed and with high accuracy.

FIG. 1 is a block diagram showing a configuration of a face image detection apparatus according to an embodiment of the present invention. FIG. 2 is a flowchart showing the operation of the face image detection apparatus. FIG. 3 is a diagram illustrating an example of original image data. FIG. 4 is a diagram illustrating an example of image data for face detection. FIG. 5 is a diagram illustrating an example of image data for face detection. FIG. 6 is a diagram illustrating an example of image data for face detection. FIG. 7 is a diagram showing an example of the appearance of a digital still camera equipped with a face image detection device. FIG. 8 is a diagram showing an example of the configuration of a digital still camera equipped with a face image detection device.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Face image detection apparatus 101 A / D converter 102 CPU
DESCRIPTION OF SYMBOLS 103 Luminance signal extraction part 104 Reduction rotation image data preparation part 105 Memory 106 Face detection part 107 CPU interface part 108 Reduction rotation image data specific part 109 Angle setting value selection part 110 Face image position determination part 111 Face image detection part

Claims (12)

A face image detection device for detecting a position of a face image using input image data,
The input image data is reduced to create n pieces of reduced rotated image data every 360 / n degrees rotated by 360 × i / n (n is an integer of 2 or more: i = 0 to n−1) degrees. Reduced rotation image data creation means;
A face image detection apparatus comprising: a face detection unit that detects a position of a face image with respect to the n pieces of reduced rotation image data. further,
Reduced rotation image data specifying means for specifying reduced rotation image data in which the position of the face image is detected by the face detection means among the n reduced rotation image data;
Face image position determination for determining the position of the face image in the input image data from the position of the face image detected by the face detection means and the correspondence relationship between the positions of the input image data and the reduced rotated image data The face image detection apparatus according to claim 1, further comprising: means. The face detection means detects the position of the face image with respect to the n pieces of reduced rotation image data using a technique capable of detecting the position of a face image having an inclination within a range of at least 360 / n degrees. The face image detection apparatus according to claim 1. further,
Reduced rotation image data specifying means for specifying reduced rotation image data in which the position of the face image is detected by the face detection means among the n reduced rotation image data;
Based on the rotation angle θ of the reduced rotated image data specified by the reduced rotated image data specifying means with respect to the input image data, out of n angle setting values every 360 / n degrees, (−θ) Angle setting value selection means for selecting one angle setting value used for detecting the position of a face image having an inclination,
The face detection means further detects a position of a face image with respect to the input image data using the one angle setting value selected by the angle setting value selection means. The face image detection apparatus described in 1. The angle setting value selection unit further includes a plurality of rotation angles of the plurality of reduced rotation image data with respect to the input image data when there are a plurality of the reduced rotation image data specified by the reduced rotation image data specifying unit. Based on θ, a plurality of angle setting values used for position detection of a face image having an inclination of (−θ) are selected from n angle setting values every 360 / n degrees,
The face detection unit may further include, when there are a plurality of the reduced rotated image data specified by the reduced rotated image data specifying unit, from among the plurality of angle setting values selected by the angle setting value selecting unit. 5. The face image according to claim 4, wherein one angle setting value is sequentially extracted, and the position of the face image is detected with respect to the input image data, using the extracted angle setting value. Detection device. The face detecting means detects the position of the face image again with respect to the n pieces of reduced rotated image data when the reduced rotated image data is not specified by the reduced rotated image data specifying means. The face image detection apparatus according to claim 4. The face detection means determines the position of the face image with respect to the n pieces of reduced rotation image data using one predetermined angle setting value used for position detection of a face image having a predetermined angle of inclination. Detect
The face image detection device further includes reduced rotation image data specifying means for specifying reduced rotation image data in which the position of the face image is detected by the face detection means among the n reduced rotation image data.
The face detection means further uses the predetermined one angle setting value when there are a plurality of the reduced rotated image data specified by the reduced rotated image data specifying means. The face image detection apparatus according to claim 1, wherein the position of the face image is detected with respect to the plurality of reduced rotation image data specified by the specifying means. The face detecting means enlarges the plurality of reduced rotated image data specified by the reduced rotated image data specifying means when there are a plurality of the reduced rotated image data specified by the reduced rotated image data specifying means. 8. The face image detection according to claim 7, wherein a position of a face image is detected with respect to the plurality of reduced rotated image data enlarged using the one predetermined angle setting value. apparatus. The face detection means determines the position of the face image with respect to the n pieces of reduced rotation image data using one predetermined angle setting value used for position detection of a face image having a predetermined angle of inclination. The face image detecting device according to claim 1, wherein the face image detecting device is detected. The reduced rotated image data creating means further creates one face detection image including the n reduced rotated image data,
The face image detection apparatus according to claim 1, wherein the face detection unit detects a position of a face image with respect to the one face detection image. An imaging device for imaging a subject,
Photographing means for photographing a subject and outputting image data;
The input image data photographed by the photographing means is reduced, and reduced by 360 / n degrees every 360 × i / n (n is an integer of 2 or more: i = 0 to n−1) degrees. Reduced rotation image data creation means for creating rotation image data;
An imaging apparatus comprising: a face detection unit that detects a position of a face image with respect to the n pieces of reduced rotation image data. A face image detection method for detecting a position of a face image using input image data,
The input image data is reduced to create n pieces of reduced rotated image data every 360 / n degrees rotated by 360 × i / n (n is an integer of 2 or more: i = 0 to n−1) degrees. Reduced rotation image data creation step;
A face detection step of detecting a position of the face image with respect to the n pieces of reduced rotation image data.

Images