An image filtering process based on foveal mechanism simulation

This paper presents a space-variant filtering process which uses a fundamental feature of the human visual system: the selective analysis of visual information. Our approach combines two fields of research generally disconnected: human vision and computer vision. An incursion in the organization of the human retina, the first neutral tissue involved in vision, associated with classical tools of image processing is at the basis of our filtering method. The resources of the vision system are focused only on the part of the studied images where information may be important for the task at hand. Images are then analyzed with a space-variant resolution and data are filtered according to the characteristics to be pointed out. The filtering process is then adaptive and can be tuned with relevant information conveyed by the scenes under study.