A Multilayer Model for Rapid Object Recognition Based on Human Visual System

Humans can effectively recognize objects in complex natural scenes within a fraction of a second. This dominant ability has motivated many computational object recognition models. The human visual system hierarchically recognizes object in several processing layers. Along these layers a set of features is extracted by different part of visual system. The different parts of the human visual system have been expertise to specific calculations and perform this task without redundancy information. In this process, the activation of neurons, a threshold is present. This threshold transmits useful and necessary information through cortex processing stream. Each proposed model should be optimized in terms of overlapping computation. In this paper, we improved a hierarchical model, and used the threshold step and saving steps in proposed model. This would reduce the redundancy and overlapping computation in the model. On the other hand, with reducing computational errors, while the speed of recognition is increased, the recognition rate has reached a desired value

Keywords:

image compression , multiple model , object recognition , receptive field threshold