Modified Methods of Generating Saliency Maps Based on

Various types of approaches that can model a human vi-sual attention have been already proposed. However, a model that could perfectly simulate the human perception and methods of computer prediction of human visual at-tention belongs to one of the high focused research areas. Our work is aimed at methods of generating a saliency map which will detect the areas in the picture that could most likely attract a human attention. The basis of our work is method of saliency detection using superpixels published by authors Z. Liu, O. Meur and S. Luo. Sev-eral modifications of this method with the aim to improve the results have been proposed, tested and evaluated in our experiments. In this paper, we present our proposed mod-ification based on border prior and statistical evaluation of the saliency central position in the used dataset. This center position will be expressed using a fitted Gaussian function. Results of all experiments are evaluated and pre-sented in the following paper