Reducing the Cost of the Hybrid Evolutionary Algorithm with Image Local Response in Electronic Imaging

Abstract

The paper focuses on the efficiency of the hybrid evolutionary algorithm (HEA) for solving the global optimization problem arising in electronic imaging. The particular version of the algorithm utilizes image local response (ILR), in order to reduce the computational cost. ILR is defined as the variation of fitness function due to a small variation of the parameters, and is computed over a small area. ILR is utilized at different stages of the algorithm. At the preprocessing stage, it reduces the area of the image participating in fitness evaluation. The correlation in the response space identifies the set of subregions that can contain the correct match between the images. Response values are used to adaptively control local search with the Downhill simplex method by applying a contraction transformation to the vector of the standard simplex coefficients. The computational experiments with 2D-grayscale images provide the experimental support of the ILR model.