Improved Facial Feature Points Calibration Algorithm Based on ASM

Abstract

The matching results of traditional ASM are greatly affected by the model initial position. An improper position will lead algorithm to fail. To enhance the accuracy of feature detection, a Quick Rough Positioning Model(QRPM) algorithm is proposed, which makes use of the image and model gray information to calculate the similar coefficient between the model and the region detected. Coarse and fine explorations are adopted to extract the rough region, where the initial model will be set. Experimental results show that the improved algorithm can effectively enhance the accuracy in facial feature points' calibration, and avoid the ASM results falling into the local minimum.