Abstract: Feature extraction is a key position in the field of image analysis,understanding and recognition.At the present stage,the approach of feature extraction ignores its own visual information of image．Based on that,feature extraction algorithms and applications based on the complex visual information of image itself was proposed.The algorithm analyzes the complex cells function of at the area V1layer 4B of visual cortex,reveals the ability of the complexity of the visual cells,i.e.extracting image region independent,non-linear and translation invariant features,creats the mathematical model of complex visual cells,and extracts complex visual information of region image.It solves the shortcomings of ignoring the visual characteristics of the image itself and achieves good search results in the field of content-based image retrieval.

Key words: Visual perception theory,Geometric invariance,Nonlinear principal component analysis,Feature extraction