Studying human visual attention become important in more and more professional fields ranging from advertising industry to surveillance system. Several attempts have been made occurred to predict the human visual attention. There were first the feature-driven topographic maps, which were used to represent the symmetries and contrast features of a picture. Nowadays eye-tracking is the state-of-art method to identify the attentional biases. Experiences reveal, that visual attention is rather influenced by meaningful ob-jects based on image semantic, than the symmetry or contrast features of the picture. That is why the separation of objects is important, which means grouping pixels for a two dimensional image.
In my thesis I will show the different kinds and the mathematical background of sali-ency maps. The technology and performance of eye-tracking will be mentioned. I will name some databases, which came into existence by this method. I will show the clus-tering methods, and then I will have a review about my software, which highlights the fixation points in the picture, the concepts derived from these points, and the visually important objects calculated by the previously mentioned things. As an addition to my project I will describe the face detection algorithm, which is used as prediction in my application implemented with MATLAB.