Face Alignment Method Based on Mouth Constrain

Mouth shape detection can effectively solve the problem of complex environments such as semi-occlusion encountered in the face key point detection. In this paper, a method based on feature point detection in the local area of the mouth is proposed. By determining 20 feature points in the mouth area and then using the regression algorithm, the position of the feature points in the mouth area is accurately determined, and then matched and predicted the feature points of the face are located. according to the distance and other characteristics. The feature points of the entire face are positioned to achieve the effect of face alignment. In this paper, the improved ESR algorithm effectively solves the problem of local feature point localization. At the same time, since we only locate the feature points of the mouth area and then calculate other feature points, we can improve the detection efficiency.