Modeling Information Processing Mechanisms of Low-level Biological Vision

E. Diamant (Israel)


image information content, image analysis, image description.


Despite impressive advances in computer technologies, contemporary human-made image processing systems remain far behind their biological counterparts. To overcome this subordinate state, we propose to take advantage of some principles borrowed from the field of biological vision and not recognized yet in today’s engineering systems. More specifically, we propose to embrace a concept of a pixel’s receptive field and its derivation - a concept of information content of a single pixel. Both of them are supposed to mock-up some known eye’s retina properties that are commonly used to model early information processing mechanisms of biological vision. We provide a simple yet efficient method for quantitative estimation of pixel’s information content. On this ground, a vast majority of image processing tasks, that require information related assumptions for their successful fulfillment, could be reformulated and redesigned to gain real performance enhancement. Examples of such tasks are provided, as well as Lena images demonstrating simulation results of new image processing algorithms.

Important Links:

Go Back