SEMI-AUTOMATED FEATURE EXTRACTION USING SIMULATED ANNEALING

An efficient semi-automatic feature extraction technique that integrates the principles of the active contour models (Kass et al 1988) with simulated annealing (Kirkpatrik et al 1983) is proposed in this article. The methodology in its initial phase requires the application of appropriate image processing techniques, computation of the energy image and a rough delineation of the feature of interest by the user. Subsequently, B-splines are used to model the initial user delineated points and a window containing both the spline and the feature of interest is considered. The searching capability of simulated annealing is then exploited for evolving a minimum energy state in a rectangular window around the delineation for extracting the feature appropriately. Comparison of the performance of the proposed algorithm with snakes, another common feature extraction technique based on energy minimisation, demonstrates the superiority of the former for situations where the initial delineation does not closely approximate the feature of interest.