Study Of Algorithm Of Face Detection Based On A Single Image

Face detection is an important research direction in the field of computer vision.It has in occupied an important position and widely used value in the socialproduction and daily life. This subject in order to solve several key issues currentlyface detection technology proposes an face detection method based on the colorsimilarity of the YCbCr color space with dynamic threshold combination of colorsegmentation, geometric rules and template matching eye location, on the basis ofresearch results of face detection. The work of this paper include the following threeaspects:(1) Color segmentationComparison of several commonly used color space, select the YCbCr color spaceto create the color model, because of the YCbCr color space is widely used, colorclustering, good features. A method for skin segmentation based on similarity of skincolor and dynamic threshold is proposed to improve face detection accuracy anddetection speed. Firstly, the similarity of skin color of the images is calculated inYCbCr space, combined with the between-cluster variance and within-class scatter tostrike a dynamic threshold, and then through the threshold value for skin colorsegmentation. the proposed algorithm improves the performance of skin segmentation，and can segment skin regions exactly in the complex background.(2) Face DetectionIn this article, the method of face detection is the Euler number and eye locationmethod. In order to eliminate the noise need first filtered the binary image obtained bycolor segmentation through morphological filtering processing. Its connectedcomponent analysis, Determine the face region using Geometric characteristics of thearea of the face, center, high aspect ratio, etc. Euler number to be screened again todetermine the face region. Last, in order to screening and verification of the possible face regionthrough the eye location method validation that based on the rules of geometry and templatematching eye location. This approach combines several face detection algorithm, fromcoarse to fine multi-layer filter to overcome the defects of a way to effectively reducethe false detection rate of the face region. (3) Three-dimensional reconstruction of the faceThree-dimensional reconstruction of face region of the face detection throughthree-dimensional reconstruction algorithm based on shape from shading.