Creation of a panoramic image by genetic algorithm

Abstract

This paper proposes a way to improve the performance of panoramic image generation by template matching using genetic algorithm (GA). GA is used to efficiently locate a template image for an object one, and can generate a panoramic image with high speed and accuracy than traditional image composition methods. Given x and y coordinates to chromosomes of genetic algorithm, GA carries out an optimization of the coordinates for genetic template matching. The present method detects the coordinates that exceeds a certain threshold value of a fitness function in template matching using luminance values. The template matching is carried out again at 8 vicinities of the coordinates to search an optimal solution. Using the optimal solution to the point that has the highest fitness value, we perform image composition of the template and the object images with respect to the coordinates. We obtained good results for specific images in experiments. However, in the case of other images, matching accuracy was worse by shear and tilt remained after the genetic template matching. It is expected an approach by using feature values improves in accuracy rather than luminance values in the future.