基于主导拮抗抑制的多尺度边缘检测

Multi-scale Edge Detection with Dominating Opponent Inhibition

图像边缘检测算子的抑噪能力和定位精度是一对矛盾,从视觉神经生理学角度出发,提出一种基于主导拮抗抑制的多尺度边缘检测模型。该模型将主导拮抗抑制机制拓展到多尺度处理,通过引入侧膝体拮抗通道的强抑制机制达到自适应抑噪的效果,获得的对比度信息经过简单细胞子域的非线性整合得到锐化的边缘输出,最后将多个尺度的处理结果进行融合,获得最终的边界轮廓。理论和实验证明通过主导拮抗抑制机制和多尺度处理的结合,获得抑噪能力强、定位精度高的边缘检测模型,该模型具有生理学基础,结构简单,计算效率高。

Noise suppression conflicts with high accuracy in image edge detection.A multi-scale edge detection model with dominating opponent inhibition was proposed from physiological point of view.The presented model made extension of dominating opponent inhibition mechanism to processing at multiple scales.Image noise can be adaptively suppressed by dominating inhibition from the opponent path of lateral geniculate nucleus.Sharp edges were obtained through the non-linear combination of simple cell subfields.Finally multi-scale processing results were combined to obtain boundary contour.Theoretical analysis and experimental results indicate that the proposed model performs well in noise suppression and accurate edge detection through the combination of dominating opponent inhibition and multi-scale processing.Meanwhile,it is biologically plausible,simple and efficient.