基于新模糊准则与 DE算法的红外人体图像分割 *

人体目标的抽取是红外人体图像处理的基础 ,为了有效获取红外图像中的人体目标 ,提出了一种新的图像阈值化方法。首先在对模糊熵方法本质分析的基础上 ,通过参数变换定义了一种新的最小化模糊准则 ,然后将差分演化 ( differential evolution, DE)算法与新模糊准则相结合用于快速、有效地获取最佳阈值 ,最后在真实红外人体图像上与其他方法进行了对比实验。实验结果表明 ,所提出的方法不仅能得到理想的分割结果 ,而且 CPU耗时也较少 ,满足实时性处理要求。

Infrared human image segmentation based on new fuzzy criterion and DE algorithm

The extraction of human target is the foundation of infrared human image processing. In order to obtain the human target from infrared image effectively, this paper proposed a novel thresholding method. Firstly, defined a new minimum fuzzy criterion through parameter transformation based on the analysis of the essence of maximum fuzzy entropy method. Then, com-bined the new criterion with differential evolution algorithm for obtain optimal threshold. Finally, tested the new method on real infrared images with other methods. Experimental results show that the proposed method not only can obtain ideal segmentation results but also the CPU time-consuming is lower, and it is suitable for real time application.