Complete fuzzy LDA algorithm in image segmentation

This paper proposes a novel method, called complete fuzzy LDA (CFLDA), which combines the linear discriminant analysis (LDA) and fuzzy set theory. LDA preserve the total variance by maximizing the trace of feature variance, but LDA cannot preserve local information due to pursuing maximal variance. So, the complete fuzzy linear discriminant analysis (CFLDA) algorithm is proposed, in which the fuzzy k-nearest neighbor (FKNN) is implemented to achieve the distribution local information of original samples. Experimental results on ORL, Yale, and AR face databases show the effectiveness of the proposed method. Image segmentation experimental results show better distinguished results in images.