Research Method Of Face Recognition Based On Spares Representation

Face recognition, which relates to many subject fields such as pattern recognition,digital image processing, artificial intelligence and machine vision, is the most widelyused in biometric technology. Face recognition identifies different people throughautomatically extracting and analysing the features of face images, it also has a broadapplication in the security area, entrance guard system and attendance system, etc.In the existing face recognition algorithms, face recognition based on sparserepresentation is a new recognition method using the discriminant of sparserepresentation. Sparse representation has a unique advantage in classification, andmakes feature selection no longer necessary conditions. In addition，this method notonly has a high recognition rate, but also has good robustness, so it has attracted muchattention in recent years.In this paper, the main research results are as follows:1The improvement technology of inferior image is researched. The purpose ofimproving the quality of image is to enhance the image resolution of which is not clear.This means that high-resolution image is reconstructed from the low resolution ofimage. An image reconstruction method based on Principal ComponentsAnalysis(PCA) transform is proposed in this paper, which can improve the lowresolution image and make the image clearer and make the whole calculation processsimpler.2The influence of sparse representation parameters on face recognition is studied.From the basic theory of sparse representation, we can know that sparse representationhas a strong natural discrimination, which chooses the most compact atoms of theinput signal and keep away from other untight atoms. Here, this capability is used inface recognition, and the experiments are made on ORL and YaleB face database, andthe result shows that the classification of the test image is determined by theparameters of sparse representation. 3Face recognition combing module PCA with sparse representation isresearched. Firstly, face image was divided into blocks, and the features of the blockimage are extracted by PCA. Then, the subimage from PCA transform will be used asa new training sample to be trained, what’s more, the atom library is represented bythis new sample one. Finally, the test images corresponding to the train set aresearched automatically via the principle of block matching, and then they are classifiedby sparse representation. On standard face database, the experiment show that theproposed method can achieve a better result.