Method and apparatus for detecting best shot in long-range face recognition system
A method and apparatus for detecting the best shot in a long-range face recognition system is provided. The method of detecting the best shot includes detecting a facial area in an image received from the outside, calculating a quality element measurement value of the face image by analyzing the facial area, and selecting a best shot face image, among the detected facial areas, based on the quality element measurement value.
1. A best shot detection apparatus, comprising:
a face image detection module configured to detect facial areas in an image;
a quality element measurement module configured to calculate quality element measurement values of the detected facial areas by analyzing the detected facial areas; and
a best shot selection module configured to select best shot face images, among the detected facial areas, based on the quality element measurement values,
wherein the quality element measurement values comprise a combination of a resolution value, a focus element measurement value, a pose element measurement value, and an illumination element measurement value,
wherein the quality element measurement module calculates the focus element measurement value by analyzing a high frequency component of the facial area using wavelet transform,
wherein the quality element measurement module calculates the pose element measurement value by analyzing a difference between a front face and three-dimensional angle of the facial area using any one of a 3D Morphable Model (3DMM) and a 3D Appearance Active Model (3D AAM),
wherein the best shot selection module calculates quality indexes of the detected facial areas based on the quality element measurement values and selects the best shot face images, among the detected facial areas based on the quality indexes, and transmits the best shot face image to a face recognition apparatus,
wherein the best shot selection module limits the number of images to be sent to the face recognition apparatus based on the number of the best shot face images.
2. The best shot detection apparatus of claim 1 , wherein the best shot selection module determines whether to transmit the best shot face images to the face recognition apparatus based on the quality indexes of the best shot face images.
3. The best shot detection apparatus of claim 1 , wherein the quality element measurement module calculates the illumination element measurement value through an analysis of a brightness histogram of the facial area.
4. A face recognition system, comprising:
the best shot detection apparatus according to claim 1 ; and
the face recognition apparatus configured to perform face recognition based on the best shot face images.
5. A method of detecting a best shot, comprising:
detecting face images in an image received from an outside;
calculating quality element measurement values of the detected face images by analyzing the detected face images; and
selecting the best shot face images, among the detected face images, based on the quality element measurement value,
wherein the quality element measurement values comprise a combination of a resolution value, a focus element measurement value, a pose element measurement value, and an illumination element measurement value,
wherein the calculating of the quality element measurement values comprises calculating the focus element measurement value by analyzing a high frequency component of the face image using wavelet transform,
wherein the calculating of the quality element measurement values comprises calculating the pose element measurement value by analyzing a difference between a front face and three-dimensional angle of the face image using any one of a 3D Morphable Model (3DMM) and a 3D Appearance Active Model (3D AAM),
wherein the selecting the best shot images comprises calculating quality indexes of the detected facial areas based on the quality element measurement values and selects the best shot face images, among the detected face images based on the quality indexes, and transmits the best shot face images to a face recognition apparatus,
wherein the selecting the best shot images comprises limiting the number of images to be sent to the face recognition apparatus based on the number of the best shot face images.
6. The method of claim 5 , wherein the selecting of the best shot face images comprises determining whether to transmit the best shot face images to the face recognition apparatus based on the quality indexes of the best shot face images.
7. The method of claim 5 , wherein the calculating of the quality element measurement values comprises calculating the illumination element measurement value by analyzing a brightness histogram of the face image.
8. The method of claim 5 , wherein the detecting of the face image comprises:
generating the detected face images by detecting facial areas in the image received from the outside, and
identifying the detected face images of an identical person based on results of a tracking of the facial area detected in a video image when the image received from the outside is the video image.