红外热图像分割中的二维灰度直方图小波变换方法

Wavelet Transformation Method of Two Dimensional Gray Histogram in Infrared Images Segmentation

利用小波分析在时域和频域具有良好的局部化特性以及二维灰度直方图所反映的邻域空间相关信息 ,提出一种基于二维灰度直方图最佳一维投影的B样条小波变换红外热图象分割方法。实验结果表明 ,这种方法利用二维信息 ,实行一维搜索 ,增强了算法对噪声的抗干扰能力。同时结合小波变换的多分辨理论 ,有效克服了红外热图像存在的图像模糊、噪声过大的问题 ,在不同的尺度之下自动选取门限阈值 ,其分割效果优于传统的分割方法。

We put forward the Cardinal B Spine Function wavelet transformation of the two dimensional gray histogram's best one dimensional projection in infrared images It takes advantages of the excellent local resolution of temporal and frequency region and the correlative neighborhood information reflected by the two dimensional gray histogram. The method improves the repellent ability of the algorithm for noise by using two dimensional information and one dimensional search. For the benefit of wavelet transform that provides the feature of multi resolution, it can overcome the infrared images'defects of obscure pictures and large noise. It realizes automatically determine thresholds and obtains excellent experimental results.