2018 | OriginalPaper | Buchkapitel
A Novel Algorithm Related with Customers Based on Image Gradient Orientation
Erschienen in:
Transactions on Edutainment XIV
Abstract
Based on the study of image gradient orientation and relevant technique about customers, this paper has proposed a algorithm related with customers based on image gradient orientation (CS-IGO-LDA). Face images are vulnerable to illumination changes, resulting in most of the traditional subspace learning algorithms which rely on image representation information are robust. In order to alleviate this problem, we represent the original samples by using image gradient orientation rather than the pixel intensity. And, in order to better describe the differences between different categories, we use methods related with customers to extract sample feature vector of each individual. The proposed CS-IGO-LDA method has made full use of the advantage of image gradient orientation and methods related with customers in face recognition. Experiments in face databases of Yale, JAFFE and XM2VTS have proved the validity of the new algorithm in face recognition and face verification.
