Multi-scale patch fuzzy decision for face recognition with category information

Small size sample face recognition is one of the most challenging problems in image classification. Multi-scale patch collaborative representation is an effective method to deal with this problem. The existing methods only consider the sample information at different scales, ignoring category information in the process of multi-scale information fusion. However, different categories of images contain different information, which has a great impact on face recognition results. To solve this problem, this paper proposes a multi-scale patch fuzzy decision method for face recognition with category information, which considers the influence of different category information on image classification results. Firstly, the fuzzy decision matrix is introduced to describe the degree of samples belonging to different categories at a single scale. Then, a weight vector is introduced for each category to describe the importance of patch scales in the category. Specifically, each element of a weight vector represents the weight value of each scale in the corresponding category, so that each category could get a weight vector at all scales. Finally, the optimization objective function for face recognition is constructed by considering the classification information both from the same category and from different categories. Experimental results on six face databases showed that the proposed method is robust and superior to most advanced small-size sample face recognition methods.