The research of chin contour in fronto-parallel images

In fronto-parallel images, chin contour is a relative stable shape feature. But it is hard to extract by conventional edge-detection methods due to the complexities of grayscale distribution in chin area. In this paper, we present a novel algorithm for chin contour extraction and a classification method of chin contour. Using the algorithm based on prior distribution and local decision, we can get the possible points of chin contour. The false points are deleted by filtering and the other points are joined to get chin contour by curve approximation. Then we do classification by analyzing the features of contours. Experimental results show that, by this algorithm, we can use chin contour as a new feature to preserve more details in face recognition. The classification result is good enough to improve the recognition rate and speed on huge face database.