A hierarchical computational model of visual attention using multi-layer analysis

Computational model of visual attention is very useful in image processing. It can improve the efficiency and save computational resource. A hierarchical computational model of visual attention based on multi-layer analysis is proposed in this paper. This model simulates the human visual attention mechanism using multi-layer data competition. Firstly, different parts of the input image compete for attention at the first layer with coarsest resolution based on their visual saliency. When a region gets the focus of attention, the sub-regions of it compete for attention at the next layer with a finer resolution. This process is done iteratively until every salient region and its sub-regions have been processed at the final layer with the finest resolution. The proposed model was tested on many natural images. Experimental results and analysis are presented in this paper. The proposed model is valid and the attention results are consistent with human visual system.