Research On 2D Facial Feature Location

With the progress of modern life and enhanced safety awareness, mora and more filed need to identify people's identity. the traditional authentication methods include iris recognition, finger print recognition, Palm print recognition. comparing with these traditional methods, face recognition have direct, friendly and convenient characteristics, therefore automatic human face recognition becomes one new hot point of identity verification and has extensive application foreground. general speaking, an automatic face recognition system contains: face detection, feature points extraction and recognition. there are some factors which affect instantiation of face recognition. they lie on pose, light and expression of face. precise facial feature points extraction is the key to solve these problems.Although people can position out the face feature point quickly and accurately,it is not easy for a computer. these difficult are mainly embodied in the complexity of people face their own, human face's posture and the diversity of environment where face is.Accurate facial feature alignment is the prerequisite of a face recognition system. currently, the Active Shape Model (ASM) and Active Appearance Model (AAM) are the main models for this problem. ASM carry out statistical analysis of the shape of the face concentrated on training date and take full account of the overall shape. when searching facial feature points, take full advantage of this face shape Priori estimates. AAM is a widely used method of quickly and efficiently based on ASM, AAM -based facial feature Point positioning method takes into account not only local feature information but also the overall shape and texture information. thougth the statistical analysis of Human face's shape features and texture features , set up the corresponding AAM model.when measure Human face's feature point , using the strategy of combination comparison adjustment recombination re-comparison , often they can be able to achieve rapid and accurate facial feature point positioning.AAM algorithm achieve relatively stable statistical face information through the texture and shape modeling, and then use face alignment method to achieve the positioning of facial features.face alignment is one core step to achieve AAM algorithm. face alignment method is optimized for the use of extreme value approach to the texture difference between the model face and the face in picture. and achieve the ultimate aim of face alignment. it can expressed by the formula:T(x_i) and T_model(x_i)) is the corresponding image pixel gray value of the real picture and AAM model picture. and it is the main way to improve the computational efficiency using different face alignment methods or calculation methods.Also , we can see though the process of AAM modeling. the accuracy of the stability of the algorithm is restricted by the diversity of the samples. generally , the more diversity of the samples, the more stability of the model. clearly,, the the stability of the model to light is limited to the diversity of model in different light illumination. in this dissertation, algorithms of facial feature points positioning as the ASM and AAM were profoundly researched. however, two models mentioned above are sensitive to the illumination variation. to fight with the disadvantages of two models, we propose an novel improvement. and the experiment shows that the improvement work well.In this paper, the main job done in this paper as follows:1. Mainly introduce the AAM algorithm widely used in the application of face recognition, and for its development in recently decade. Give two detailed representative algorithms basic AAM and inverse composition AAM, followed the detailed explanation of the two.2.At present in the majority of face recognition algorithms, light shadow removal problem has always been a hot one. this paper presents a theory of light invariant. we can easy remove the shadow of objects under illumination though the theory.3.There exist difficult in face feature positioning. such as the the complexity of people face expression, diversity of people face. although AAM is sensitive to the illumination variation, to fight with the disadvantages of two models, we propose an novel improvement. combination with this algorithm,we give a group of face picture under different illumination (not include in the samples). compared with the traditional AAM algorithm. the experiment result work well.