Subjective contour generation using a pixel-parallel anisotropic diffusion algorithm

Abstract Subjective contours are imaginary contours that the vision system in the human brain generates to complete the lack of image information. In order to generate subjective contours, we propose an anisotropic diffusion algorithm. The proposed algorithm consists of two parts: a directional-diffusion algorithm (DDA) that generates analog ridge-like functions of the diffusion state and a ridgeline detection algorithm (RDA) that generates binary line-segments by tracing the ridgelines. We show simulation results of subjective contours generated by the proposed algorithm. Since these algorithms can achieve a pixel-parallel operation based on the framework of cellular neural networks and cellular automata, they are suitable for the LSI realization.