The Color Face Image Recognition Of Combining Weighted Color Correlogram And Improved LBP Method

Since 90’s in last century, study of face recognition technology is the research hot-spot in the field of pattern recognition. Image source used in the most of existing face recognition technology is gray-scale with 256 kinds of gray level,while image collected in practical application is color. Therefore, domestic and foreign research institutions or technology companies began to study color face recognition methods,and a lot of methods have been proposed. We also do some studies on the color face recognition methods as following:1.research of the weighted color correlogramFirst, we analyzed the color space models of the color image. Then, we introduced two kinds of color feature extraction method--color histogram and color correlogram. Finally, we proposed a novel color feature extraction method-- the weighted color correlogram. The method use different color of the color image into a one-dimensional feature vector through uncertaion weight distribution based on the color-correlogram. The experiment operated on the color face library of California institute of technology. our method applied to color face recognition has higher identification accuracy compared with color correlogram and color histogram.2.research for improved LBP methodwe do some reaches on LBP, and proposed improvement of LBP according to the insufficient of applying LBP to face recognition. We obtained neighbor radius through self-adaption, and adopt threshold setting in encoding process. We oper ated our experiment on the color face library of California institute of technology,result indicated that the improved LBP applied to color face recognition has higher ide ntification accuracy.3.research on combine the weighted color correlogram with the improved LBPThe weighted color correlogram reflect the overall color distribution in the color face image, but also reflect the spatial correlation among colors and the distribution of color tightness. The improved LBP indicated the texture features of the gray image by self-adaption obtained neighborradius and threshold setting. We adopted effective method combines both the advantages to implement color face recognition efficiently.The results show that combination of the weighted color correlogram and theimproved LBP has certain improvement in accuracy and color histogram, the weighted color correlogram has achieved higher recognition rate. For color face recognition performance is improved.