基于惯性矩的图像融合算法  

Image-Fusion Algorithm Based on Inertia Quadrature

作　　者：陈彦燕[1] 王元庆[1] 

机构地区：[1]南京大学电子科学与工程系,江苏南京210093

出　　处：《计算机测量与控制》2009年第6期1190-1191,1194,共3页Computer Measurement &Control

基　　金：国家自然科学基金项目(60472026);江苏省高技术研究计划项目(BG2004031);国家高技术研究发展计划项目("863计划")(2005AA311020)

摘　　要：惯性矩作为图像纹理分析的特征量,能有效反映图像的清晰度,据此提出一种新的图像融合算法;该方法首先将源图像分割若干块,并从水平、垂直、+45°、-45°4个方向分别计算出各图像块区域的惯性矩,然后根据惯性矩和清晰度间的关系,对图像作融合处理;融合图像质量除信息熵、对比度等常用方法外,还采用结构相似度法进行客观评价;实验对比结果表明,基于惯性矩法得到的图像清晰度、对比度等均有提高,明显优于传统图像融合算法。Inertia Quadrature as the characteristic of image texture can effectively reflect image clarity. Based on it a new image fusion algorithm is proposed. First it divides the source images to blocks and computes the inertia quadrature of image blocks from four directions.. horizon, vertical, ＋45°, -45°, and then according to the relationship of inertia quadrature and image clarity, fuses images for the required one. Besides common objective rules such as entropy, contrast and etc, structure similarity is also adopted to judge image quality. The experiment result shows that clarity and contrast of the fused image are improved, apparently better than traditional methods.

关 键 词：灰度共生矩阵 惯性矩 图像融合 结构相似度 

分 类 号：TP791[自动化与计算机技术—检测技术与自动化装置]