DEFORMED SHAPE RETRIEVAL BASED ON THE HIDDEN MARKOV MODEL

Abstract

A new deformed shape retrieval method based on the hidden Markov model (HMM) is proposed in this paper. Shape features as well as statisical and contextual information are incorporated into the HMM to derive probability values. Then, the probability values can be considered matching scores to retrieve similar shapes. The proposed method is translation, rotation and scale invariant. In addition, it is robust to various non-rigid deformations such as perspective, shear, occlusion and monlinear distortions. The advantages are accomplished by the strategies we adopt. First, the proposed shape features are translation, rotation and scale invariant. Secondly, th flexibility of shape matching can be increased since the HMM is high tolerance to noise and distortion. Henceforth, non-rigid deformation can be coped with. Finally, although the HMM is computation intensive, only few high-level shape features are necessary for shpae representation, thus, the computation efficiency is satisfactory. Our method has been applied on two databases: geometry and character. The experimental results prove the effectiveness, robustness and practicability of the proposed approach.