2016 National Conference on Electrical, Electronics and Biomedical Engineering, ELECO 2016, Bursa, Turkey, 1 - 03 December 2016, pp.143-146
© 2016 The Chamber of Turkish Electrical Engineers.In the presented study, training of kernel matrix which is used for convolution or correlation image filtering operations using genetic algorithms was analyzed. For this purpose, filtering performance was analyzed according to the size of kernel matrix and the size of the image. In the experimental studies, various sizes of images and various example images for each size were used. Noisy images were formed by adding Gaussian noise. Computations were obtained for fixed number of population and iterations to see the effect of parameters. According to obtained results, the effect of image size variation is at small levels when the same filtering kernel is used. Performance degradation was observed as the size of the kernel matrix is increased. This degradation which is due to the increase in the number of variables of genetic algorithm is improved by increasing the number of iteration.