在深入研究红外弱小目标的成像特性, 并分析现有分割算法的基础之上, 针对空中红外弱小目标提出一种基于局部极大值的红外弱小目标的分割方法。该方法利用红外弱小目标所在像素点与周围像素点的灰度值的梯度关系来搜索图像中的局部极大值, 并通过多步长融合表决分割红外弱小目标。仿真实验证明了算法的有效性。

A new method based on local maximum is presented for segmenting small dim infrared targets. With the analysis on properties of many infrared images and the research on existing segmentation methods, it can be found that the gray values of target pixels always present to be the maximal in the local area. Building on the principle of IR radiation property of target, the method uses the gradient variations of pixel values in both horizontal and vertical directions to get the local maximum, and then segments targets by means of multi-step gradient voting fusion. The experimental results testify the perfect efficiency of the presenting method.