Research On Image Texture Feature Extraction Algorithm Of Complete CS - LBP Operator Based On Gabor Filter

In recent years, local binary pattern (LBP) has been widely applied to image texture feature extraction, and achieved good results.Therefore, the improved local binary pattern algorithms are good methods to extract image texture features.Improved Local Binary Pattern algorithms include centrally symmetric local binary pattern (CS-LBP), completed local binary pattern (CLBP), adaptive local binary pattern (ALBP) and so on. In addition to LBP, Gabor filters have also been widely applied to image texture feature extraction.Naturally, people will think of the combination of Gabor filters and LBP applied to the field, However, taking into account the LBP algorithm code is 8, which would make a lot of calculation, the length of centrally symmetric local binary pattern(CS-LBP) code is 4, if you use CS-LBP encoding algorithm, it can reduce the amount of calculation.Given this situation, combining the Gabor filters and CS-LBP algorithm appears,Under this condition, combining the Gabor filters and CS-LBP algorithm is worthy of value, which extracts more texture information and reduce the computation complex.To solve these problems, this paper was carried out a series of studies.The main works have been done as follows:1、Although the CS-LBP algorithm can reduce the computational encoding, but the corresponding loss of some of the main information of the image.Taking into account these problems, researchers have proposed a completed center symmetric local binary patterns (CCS-LBP)algorithm for texture feature extraction, the original CS-LBP only considers the symbolic information of the image,and ignores the part of its value and the center pixel.The CCS-LBP algorithm is to extract image texture features three histogram, and then by way of integration, to retrieve a new image texture feature vectors.In this paper, some works have been done to improve the fusion, usually adopts the concatenated fusion method, but use in this way to get a large number of feature dimensions.Given this situation, this paper uses a joint integration approach, Experimental results on ORL, Yale and Valid standard face database show that the use of joint fusion approach CCS-LBP algorithm to achieve high recognition rate than using concatenated fusion approach CCS-LBP algorithm.2、As combining Gabor filters and CS-LBP algorithm to extract texture features are not enough rich and comprehensive, the paper proposed combining Gabor filters and CCS-LBP to extract texture features algorithm. Three steps have been done in our proposed algorithm.the first step is to use Gabor filters for image filtering. then after filtering characteristics of the obtained image for CCS-LBP code. finally, these images through CCS-LBP feature vectors obtained by coding, re-formation of a new texture feature vector through the fusion way. Some experiments have been done on ORL, Yale and FERET face databases. Experimental results show that the recognition rate of the combining Gabor filters and CCS-LBP algorithm is better than combining Gabor filters and CS-LBP algorithm and CCS-LBP algorithm.