Graphics and Image Pre-indexing based on Tolerance Rough Sets

All previous works on information retrieval using rough sets are based on equivalence rough sets model (ERSM), but the property of equivalence has limited its application fields. Some researchers have proposed a kind of tolerance rough sets model (TRSM) for information retrieval. The core of TRSM is tolerance classes constructed by the index co-occurrence and a matching algorithm with tolerance rough inclusions. In this paper a new method for graphics and image pre-indexing by tolerance rough sets is presented. This new approach organizes the graphics and image in an approximation space of tolerance classes. Experimental results obtained from human face graphics and image pre-indexing show that the tolerance approach can effectively overcome the limitation of ERSM and enhance the efficiency of graphics and image indexing.