基于颜色恒常的照度准不变边缘分类算法

现有的照度准不变边缘分类算法在标准光源下分类效果较好,但是该算法没有考虑光源颜色的影响,而实际上光源颜色对算法的分类效果影响很大,因此提出了一种基于颜色恒常的照度准不变边缘分类算法.该算法首先估计光源颜色,然后用估计出的光源颜色参数矫正照度准不变边缘分类算法.与传统的照度准不变边缘分类算法相比,该算法由于能估算出光源的颜色,所以当光源颜色发生改变时,矫正以后的照度准不变分类算法仍然对光照颜色改变具有鲁棒性.实验结果显示,基于颜色恒常的照度准不变边缘分类算法,在光照颜色变化的情况下,不仅能检测出更多的边缘,噪声点更少,而且检测的边缘更准确,稳定,线条更光滑连续.

Edge Classification by Photometric Quasi-invariants Based on Color Constancy

The existing algorithm of edge classification by photometric quasi-invariants has a good performance under the standard illumination.However,the algorithm doesn′t consider the effect of the color of the luminous source ,while the color of illumination in fact greatly influences the performance.This paper proposes a novel approach to color edge classification by photometric quasi-invariants based on color constancy.The color of luminous source is evaluated by the algorithm,then the algorithm of edge classification is revised by the parameter of the luminous color.Compared to the traditional algorithm of edge classification,this algorithm has robustness to the color of illumination when the color of the light source changes.Experiments show that the performance of the new approach to the edge classification is better than that of the traditional kind in the edge accuracy,stability and the edge smooth and continuousness,except for more edges and less noise.