The paper deals with the topic of image registration. Image registration is important operation in image processing. Registration accuracy has essential importance for the quality of further processing of images in e.g. 3D processing. This paper provides one new cost function. Experimental results performed in this work show that this newly proposed function has the advantage of low complexity in comparison with the present-day methods. The function has good reliability and fast convergence. The paper presents a specific modification of particle swarm optimization for the evaluation of the results as well. The paper provides a novel approach using pseudo-color representation of image. Pseudo-color representation offers high contrast in images.