Gabor Active Shape Model And Its' Application To Face Segmentation

Recently, image engineering becomes one of the best developing subjects in the world. It includes three layers, Image Processing, Image Segmentation and Image Understanding. Image segmentation is a key step from the Image Processing to Image Understanding, which is also a basic technique in the computer vision. It makes the high-level computer vision tasks possible, so it plays an important role in the technology of the image processing. Face Recognition is one of the most active research fields on pattern recognition, there are more and more applications in this field on security, user interface and artificial intelligence. In most circumstances it needs accurate face segmentation, so human face segmentation is a promising field in computer vision.Active Shape Model (ASM) is one of the popular methods in image segmentation. It bases in Point Distribution Model (PDM). It needs to train a shape and local appearance model, and then put the model initially on the human face image for segmentation. Iteratively change the model until it convergence to the edge. The traditional ASM use local appearance information to search, it is suffers from variations of pose, illumination and expressions.In this paper, we present a method, Gabor ASM, in which gabor wavelet features in every landmark are used for modeling local image structure. The magnitude and phase of gabor features contain reach information about the local structural features of face images to be aligned, and provide accurate guidance for search. To a large extent, the repairs defects in gray scale based search. A coarse to fine grained search is used to position local features in the image. First use skin similarity image search to get the rough shape then use gabor parameter to renew the rough shape. Experimental results demonstrate the ability of Gabor ASM to accurately align and locate facial features.