Face Recognition Using a Face-Only Database: A New Approach

摘要: In this paper, a coarse-to-fine, LDA-based face recognition system is proposed. Through careful implementation, we found that the databases adopted by two state-of-the-art systems[1,2] were incorrect because they mistakenly use some non-face portions for recognition. Hence, face-only database used in proposed system. Since facial organs on human only differ slightly from person to person, decision-boundary determination process tougher than it conventional approaches. Therefore, order avoid above mentioned ambiguity problem, propose retrieve closest subset of samples instead retrieving single sample. The has several advantages. First, able deal with very large and can thus provide basis efficient search. Second, due its design nature, handle defocus noise problems.Third, faster autocorrelation plus LDA approach [1] PCA [2], which are believed be statistics-based, systems. Experimental results prove method better traditional methods terms efficiency accuracy.