Multi-pose color face recognition in a complex background

Face recognition has wider application fields. In recurrent references, most of the algorithms that deal with the face recognition in the static images are with simple background, and only used for ID picture recognition. It is necessary to study the whole process of multipose face recognition in a clutter background. In this paper an automatic multi-pose face recognition system with multi-feature is proposed. It consists of several steps: face detection, detection and location of the face organs, feature extraction for recognition, recognition decision. In face detection the combination of skin-color and multi-verification which consists of the analysis of the shape, local organ features and head model is applied to improve the performance. In detection and location of the face organ feature points, with the analysis of multiple features and their projections, the combination of an iterative search with a confidence function and template matching at the candidate points is adopted to improve the performance of accuracy and speed. In feature extraction for recognition, geometry normalization based on three-point affine transform is adopted to conserve the information to a maximum content before the feature extraction of principal component analysis (PCA). In recognition derision, a hierarchical face model with the division of the face poses is introduced to reduce its retrieval space and thus to cut its time consumption. In addition, a fusion decision is applied to improve the face recognition performance. Also, pose recognition result can be got simultaneously. The new approach is applied to 420 color images which consist of multi-pose faces with two visible eyes in a complex background, and the results are satisfactory.