SENSITIVITY ANALYSIS OF GENERALIZED FINITE AUTOMATA FOR IMAGE RETRIEVAL

摘要

Finite automata have been used mainly for image and video compression in the field of image processing. This paper evaluates finite automata within the context of content based image retrieval in the compressed domain, where an automata-based representation is proposed as a way to avoid image decompression prior to retrieval. The removal of the decompression process aims to improve image matching and retrieval performance with respect to accuracy, speed and computational overheads. The paper presents a sensitivity analysis of image presentation in the automata domain. This study examines the effect of some image alterations on the corresponding automata. Experimental results show that automata yield successful image matching despite image alteration. This reinforces the case for automata-based image representation in content-based retrieval.