基于边缘信息和Otsu的红外图像分割方法

摘要

针对Otsu法对目标面积过小或灰度直方图不呈明显双峰特性的红外图像难以有效分割的问题,了提出一种使用边缘信息来改进直方图形状的方法,这种方法使得图像不受目标和背景相对大小的限制,且改善了直方图峰值的对称性,获得了被深谷分开的双峰,与传统的Otsu法相比,该方法能够取得更理想的分割效果,更好地可以提取感兴趣的红外目标.%Infrared images with small target or gray histogram without obvious double crest characteristics are hard to segment effectively .To solve this problem , a method for improving histogram shapes by edge information is proposed .This method overcomes the restriction of the relative size of the target and background and improves the symmetry of histogram crest value .And we can get double crests separated by a deep valley .Compared to traditional Otsu method , the improving method can obtain the more desirable segmentation results and get better interest target .