Tomographic reconstruction of circular and elliptical objects using genetic algorithm

This paper addresses the model-based tomographic reconstruction from a limited number of noisy projections for the detection and estimation of multiple circular and elliptical objects of known intensities placed on a known uniform background. As the direct computation of the maximum-likelihood estimate is impractical even for small number of objects in the image, we have used an optimization strategy based on the genetic algorithm (GA). The GA uses variable-length chromosome coding and two knowledge-based operators namely, add object and delete object. The proposed algorithm correctly detects the number of objects in the given images and the estimation of the object's parameters, namely, the location, size and shape is fairly accurate.