摘要
对色彩结构码(CSC)算法做了改进,将图像边界信息结合进CSC算法的连接过程,并以色度、饱和度、亮度(HSV)色彩模型为基础确定色彩差别.通过设定颜色差别和边界梯度的容忍阈值控制分割的深度.实验用中国虚拟人(VCH)和虚拟人计划(VHP)数据集的彩色切片图像进行分割.结果表明,分割结果比原算法更准确,而且克服了原算法因分割结果对容忍阈值非常敏感而难于取得合适的容忍阈值的缺点.
An improved color structure code （CSC） algorithm combining region and edge information in link phase was presented. The difference of color is calculated on hue saturation value （HSV） color model. The tolerance of color difference and edge gradient controls the depth of segmentation. The algorithm was tested on color images in visible Chinese human （VCH） and visible human project （VHP） dataset and shown effective. The tolerance is more insensitive in the improved CSC algorithm so it is easy to get the right tolerance for good segmentation.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2005年第10期1685-1688,共4页
Journal of Shanghai Jiaotong University
