Abstract: The extraction methods of present digital camouflage texture are lack of subjective visual perception. With the poor effect and quality of extraction, a novel extraction arithmetic for digital camouflage texture image based on human visual perception and isoperimetric theory is presented. The proposed method first utilizes characteristics of human visual perception to build constructor of edge weight, then uses isoperimetric ratio in isoperimetric theory based on human visual perception as a criterion to determine the optimal threshold from the candidates. Experimental results on a series of natural texture images show that authors’ method outperforms several existing classic thresholding methods in segmentation quality.

Key words: digital camouflage, isoperimetric theory, multilevel thresholding, image segmentation

摘要： 现阶段数码迷彩纹理的提取方法缺乏主观视觉感知，针对其提取效果和质量欠佳的问题，提出一种融合视觉感知和等周理论的数码迷彩纹理提取方法。此方法先利用人的视觉感知特性构造边权值函数，将等周率作为选择分割阈值的准则选取最小等周率所对应的候选阈值作为最终分割阈值，完成数码迷彩纹理的提取。在一系列自然纹理图像上的实验结果表明，与现有的几种经典多阈值分割方法相比，此方法的分割效果更好。

关键词: 数码迷彩, 等周理论, 多级阈值, 图像分割