A realistic mammalian retinal model implemented on complex cell CNN universal machine

The visual system is probably the most important sensory modality for humans as well as for mammals. Its first and best-known part is the retina, which is not a mere photoreceptor or static camera but a sophisticated feature preprocessor with a continuous input and several parallel output channels. These channels build up a "visual language" and any realistic mammalian retina model should generate the elements of this visual language. The framework of mammalian retinal modeling via multi-layer CNN has been recently published. In the present paper we show the transformation of this model into a CNN-UM algorithm and the design steps of the implementation of this complex visual language. The analogic algorithm consists of a series of different complex-cell CNN dynamics. The algorithm is feasible on a recently fabricated complex cell CNN-UM chip. The decomposition method of the multilayer mammalian retina model will be discussed in detail.