针对红外图像由于目标和背景边界模糊, 采用单一熵阈值法进行图像分割结果不理想, 提出了一种基于距离灰度补偿的红外图像增强方法, 利用距离作为空间信息对灰度进行补偿, 改善了目标和背景边界模糊对图像分割的不利影响; 然后提出了一种基于交叉熵约束的最大熵阈值图像分割方法, 在交叉熵约束保证类间差异的前提下利用类内均匀性进行图像分割, 避免了单一熵方法阈值的局限性。实验结果表明, 对小目标复杂背景和复杂目标大背景的红外图像, 所提出的方法得到了准确的图像分割结果。

Basic entropy methods do not improve on segmentation of infrared images whose target and background have fuzzy boundaries. To mitigate the impact of fuzzy boundaries, an enhancement algorithm for infrared image was proposed. It takes Euclidean distance as gray compensation. Moreover, on the basis of maximum entropy, an image segmentation method was proposed which under constraint of cross entropy. The new method considers uniformity within each category on the premise that difference between categories is ensured. Experimental results show that proposed methods decrease the noise and get accurate segmentation of two images.