二维信息熵的红外图像分割算法

为了准确实现目标分割,提出将信息熵应用于红外图像的分割算法。对一般二维直方图最大熵进行推广,给出了外接4邻域直方图最大熵、灰度级-邻域灰度级绝对差直方图最大熵,以上3种二维信息熵算法均能有效地实现红外图像分割。引入属性直方图的概念,构造合适的属性集,先缩小目标的搜索范围,在此基础上运用信息熵进行目标分割,与单纯信息熵分割算法相比,得到的分割结果图中,目标的形状比较完整且引入的干扰较少。仿真结果表明该算法是有效的。

A Research of An Infrared Image Segmentation Algorithm Based on the Two-dimensional Entropy

In order to realize segmentation of the IR (infrared) small target image accurately, this paper proposes a segmented algorithm by applying the information entropy method to the infrared image. This paper not only takes the aspect of distribution of gray information into account in the two-dimensional entropy method, but also utilizes fully the spatial neighbor information of the pixel to obtain an ideal effectiveness of segmentation. After the introduction of the maximum entropy method based on the traditional two-dimensional histogram, other two methods based on External 4-connected G-A(Gray level-Average gray level) histogram and the G-G(Gray level-Gray absolute difference) histogram are given and the above methods all work well in the IR small target segmentation. Besides, the bound set of the IR image and the corresponding bound histogram are constructed to narrow the target search scope, and based on its bound histogram the IR image is segmented, by using the above the integer target is obtained with less noise compared with the pure entropy methods. The experiment results show that the algorithm is effective.