Local Pattern For Face Recognition

Face recognition possesses crucial theoretical value as well as broad application prospects. Though the techniques of face recognition have made great progresses, the performance of the existing algorithms cannot meet the actual requirements of various applications. In some sense, the face-modeling problem lies in the key point of the problem of face recognition. In recent years, generative models based on human visual mechanism, especially the local modeling method, have obtained increasing attention in the domain of vision. Based on the analysis of those existing generative models, this paper conducts research on the local modeling based face recognition method, and obtains the following results:(1) Proposes a Densely Local Matching (DLM) based face recognition method, in which by densely sampling original images we can get several sub images, which reflect features of different local regions, and obtained the final matching similarity through syncretizing the results of local subimage matching similarity. Besides that, this paper makes a lot of experimental analysis of varying image-dividing strategy (densely, sparsely, and different size of sub blocks etc.), and syncretizing strategies (entirely syncretize and partial syncretize). The experimental results indicate that the local matching method of partial syncretizing the similarities of densely sampling sub blocks can get better performance.(2) Proposes generative model based on Local Visual Primitives (LVP) for face modeling and classification. The LVPs, as the pattern of local face region, are learnt by clustering a great number of local patches. Visually, these LVPs correspond to intuitive low-level micro visual structures very well, and they are expected to constitute those high-level semantic features, such as eyes, nose and mouth. We show that only by hundreds of these LVPs can we reconstruct the high-dimensional face images (original dimension is 56x63) very effectively. Furthermore, this paper processes a LVP based face recognition method, in which spacial histograms is constructed by certain LVPs indexes of images, and then histogram intersection is applied to implement the face recognition. Primary...