基于C-V方法改进的红外图像自动分割

Automatic segmentation of infrared images based on improved C-V method

基于简化的Mumford-Shah水平集图像分割模型,Chan-Vese提出了不依赖于图像边缘的水平集图像分割算法(C-V方法).但是该方法分割参数难以确定,对于具有非均匀灰度背景的红外目标图像常常分割失败.针对这一问题给出了改进的拟合能量模型,新模型兼顾到了目标的同质性信息与其所占面积比例的关系.基于该模型的水平集图像分割方法自适应于灰度起伏的背景,可以较为理想地分割出与背景灰度差异不太明显的目标,对小目标也具有很强的适应性.实验结果表明,在固定水平集分割参数的情况下,新方法对于不同类型、不同背景的红外图像具有了良好的适应性.

Owing to the level set image segmentation method （C-V method） proposed by Chan and Vese, the energy function used in C-V method was improved, in the new energy model, the homogeneity of a target and its area ratio to background was pay attention to. The new level set based automatic target segmentation scheme adapts to clutter background, being suitable for the segmentation of small or low contrast targets The experiments showed that the new method could be adapted to infrared targets against clutter background, and the segmenting parameters can be selected easily.