The problem of how to partition on block SRC algorithm under sunglasses occlusion

The performance of face recognition systems is affected by sunglasses occlusion and how to effectively reduce the influence is particularly important. Research shows that, by block SRC algorithm, recognition rates can be improved. However, recognition rates are changed with the total number of sub-blocks and partition methods. Therefore, in this paper, we make a study on recognition rates in different sub-blocks and different partition methods. According to the results, we get the best partition method that each facial feature is divided into one or more sub-blocks respectively. Experimental results show that, when the number of sub-blocks is 5 × 2, where each facial feature is justly divided into two sub-blocks, recognition rates without occlusion or with sunglasses occlusion are 97.5% and 96% respectively.