Wavelet-based primal vision model with selective attention mechanism

Abstract

We propose a scheme for primal vision model. In this scheme, a new spatial-dependent sampling approach based on wavelets' characteristics is proposed to represent an image with partial information and to form multi-frequency channels. A feedback channel imitating the human's "eye movement" is introduced to form a close-loop control. Finally, we attempt to realize the primal vision model with wavelet-based selective attention processing mechanism. Some experiments show that the proposed scheme is reasonable and effective.<>