基于颜色拮抗和纹理抑制的轮廓检测模型

Contour detection model based on color opponent and texture suppression

作为目标识别的关键步骤,轮廓检测已成为计算机视觉研究领域的热点之一.仿生学研究发现,在初级视皮层(V1)细胞中,驱动彩色亮度单元的双拮抗细胞感受野对亮度和颜色信息敏感且具有方向选择性,对于轮廓检测起到重要作用.本文提出一种基于颜色拮抗和纹理抑制的轮廓检测模型,通过二维高斯差分DOG函数来模拟纹理抑制模板,对不同的双拮抗细胞通道进行纹理抑制,在基于颜色拮抗特性检测模型中考虑了抑制纹理的作用.实验结果表明:在BSDS300图像库下本文模型在纹理抑制方面较现有模型有一定的优势,能够较好的提取目标轮廓,是颜色拮抗目标轮廓检测模型中的一种新思路.

As a key step in target recognition, contour detection has become one of the hotspots in the field of com- puter vision research. Bionic studies have found that in primary visual cortex （V1） cells the double-opponent cell receptive field is antagonistic to color and spatially responsive, sensitive to brightness and color information and has orientation selectivity, which plays an important role in contour detection. In this paper, a two-dimensional Gaussian difference DOG function is proposed to simulate the texture suppression template. And the texture suppression of different double antagonist cell channels is made up to make up for the defects of the traditional color antagonism algorithm in ignoring the background texture. Experiments show that under the BSDS300 image library, the algorithm considers the effect of suppressing texture, enhances the weak texture and suppresses the background texture in each group of color antagonism and the model can achieve better contour detection. It is a new idea in the color opponent target contour detection model.