Based On The Structure Of Biological Image Local Toward Characterization Model And Its Calculation Efficiency Analysis

In the past several decades, Computer Vision has rapidly developed into a particu-larly striking research field in today’s computer science research, especially in the field of Artificial Intelligence, and witnessed the emergence of extensive literature. This topic includes the application of image processing in our real lives and productions, and also the theoretical exploration of visual recognition; in either of these research subfields, the representation of image signals has always been one of the key issues that is worth the effort to deal with. And this problem is also a particular concern in this paper.With the development of Artificial Intelligence, such a multi-interdisciplinary aca-demic field, both Neurobiology and Cognitive Psychology have shed great light on the Computer Vision research. Therefore, in the design of the visual computation model, on one hand, we introduce many outstanding features in human’s early visual processing pathways like the non-uniform sampling of retina, the Differences of Gaussian (DoG) model of the receptive fields of ganglion cells, as well as the parallel processing of visual pathway, etc.; and on the other side, through the effectiveness analysis on this biologi-cal structure based visual computation model, we try to examine the implicit principles beneath some prominent biological structure and how these features come into being.The visual computation model built in this paper will detect the "linear boundary segments" of the input image which are also termed as "local orientation features" of the visual stimulus, and treat these "linear boundary segments" as the basic elements of the input image to rebuild and represent the original image. In addition, we adjust the various parameters in the model itself through experiments and conduct some data analysis on the effectiveness optimization. Last but not least, we make discussions on the conditions under which this visual computation model is able to achieve the op-timal tradeoff that "both guarantees adequate representation capability and maintains relatively low energy consumption level".