Outline

I am trying to devise a biologically motivated computer based vision system that can be capable of recognizing objects in clutter and partially occluded environments. The Space-variant visual processing, that are found in biology, strip off amount of data  extracted from visual sensor , which extensively process information in the central (foveal) region of the field-of-view and maintaining  less processing resources at peripheral regions to avoid the combinatorial explosion of information contrarily to conventional system.

This system is intended for still as well as moving images. I have investigated Log(z), Log(z+a)  Bolduc and Levine, proposed touching and overlapping receptive field model of rectilinear and hexagonal model of  Log(z) and Log(z+a) retina tessellation with detailed design of algorithm and its implementation. The space variant nature of primate’s retina has taken into account while implementing the different retina tessellation. The results on all retinas mosaic have been generated by back-projecting on the input domain. The existing self-organized retina tessellation has been shown with its response. All the retina tessellation will be validated in terms of calculating optical flow field, saliency and inhibition of return map generation from still to moving images in the future course of action. Currently, I am trying to answer which retina tessellation is best in terms of speed and accuracy on the current available system.

The human retina processes large amount of data from its direct and indirect vision simultaneously. Only required information is strip off from visual scene through retinal photoreceptors before transmitting to the visual cortex where extensive processing takes place. The evolution of biological vision systems has needed to balance these requirements of high acuity and the need for a wide field of view, while facilitating the efficient distribution of limited processing resources. Even modern computational machinery seems to be irony while dealing with vision processing tasks. If whole field of view that comprises direct and indirect visual information, were to process without data reduction, then our brain would have to be an orders of magnitude larger  and weighing at least 60kg.