基于属性直方图的快速Otsu红外目标检测算法

Fast IR target-detecting method using Otsu algorithm based on bound histogram

为了提高红外小目标检测的有效性和实时性,提出了一种改进的二维Otsu算法。该算法运用了属性直方图的概念,通过构造适合红外小目标图像特性的属性集,来确定Otsu算法的最佳分割阈值。为了减小计算复杂度,从推导递推关系式和缩小搜索范围两方面,给出了快速算法。在数学形态学Top-Hat变换对原始红外图像进行背景抑制的基础上,利用本文提出的改进算法在含有噪声的背景抑制图中分割出候选的目标点,并和其他目标分割方法进行比较,实验结果证明了这种方法的抗干扰性更好、时效性更高。

A modified two-dimensional Otsu algorithm is put forward to enhance the effectiveness and the real-time in the infrared small target detection.The bound histogram is used here to create the attribute set that is suit for the infra-red small target image,and the attribute set is used to compute the optimum threshold of the Otsu algorithm.The fast method is realized through the recursive relational expression and the contraction of the searching band,which can re-duce the complexity of operation.The modified algorithm is performed in the image segmentation to select the possible targets after the background suppression with the morphological Top-Hat operator,and the target segmentation experi-ment is done by using other methods.The results show that the modified Otsu algorithm performs better in the noise-proof feature and the real-time operation.