像素级多分辨图像融合若干问题分析及实验

针对前人的工作和我们的实验进行了理论分析,讨论了像素级多分辨图像融合的三个基本问题:①为什么基于多分辨数据结构的融合能消除来自不同图像的特征之间拼接的板块效应而过渡自然?②如何构造新的融合图像的多分辩数据结构,从而能稳定地重构融合图像?③所选择的系数是否携带着原始图像的显著特征(边缘和纹理)?给出了像素级多分辨图像融合的三个基本规律:多分辨数据融合的内在机理,融合算法的稳定性条件以及基于能量准则的数据融合原理.并以此规律为引导,开发了一种新的基于小波变换域的图像融合算法,进行了实验验证.

An Analysis of Pixel Level Multi-resolution Image Fusion

Three basic problems of pixel level multi-image fusion using multi-resolution method have been theoretically studied, which are (I)Why can the image fusion algorithms using multi-resolution methods eliminate the artificial marks? (II)How to make up a new coefficients pyramid to get stable fused image reconstruction? (III)Whether the selected coefficients represent the salient patterns in the source images? Three basic principles in multi-image fusion using multi-resolution methods: the intrinsic mechanic of this kind of image fusion, the condition of a stable image fusion and the principle of the energy rule are obtained. Follow up these principles, a new algorithm using wavelet transform is developed and the experiments is performed using multi-spectral images.