A contour perception model that simulates the complex connection pattern of the visual cortex

Contour detection is the basic content of image processing and plays an important role in image analysis and target recognition. This paper proposed a contour perception model that simulates the complex connection pattern of the visual cortex. The connection included the feedforward input from the lateral geniculate body (LGN), the horizontal input from the neurons in the same layer, and the feedback input from the advanced visual cortex. Using the sparse coding characteristics of the LGN, the windmill-like structure receptive field of the primary visual cortex, and the hue perception characteristics of the advanced visual cortex to improve the accuracy of the contour extracted by the proposed model. Choosing the BSDS500 natural scene dataset as the experimental object, the F-score is selected as the evaluation index. The average optimal F-score of the proposed method is 0.72, which is better than other mainstream biological vision-based methods. Concurrently, the NYUD dataset is used for further verification. To comprehensively verify the effectiveness of the model proposed in this paper, Performance-value rather than F-score is selected as the evaluation index. The average optimal Performance-value of the proposed method is 0.42, which shows better results, too. The complex connection pattern allows neural encoding and decoding to make full use of the characteristics of information exchange between the visual cortexes, which is more in line with the biological vision system.