基于框架提升变换的多源图像融合研究

项目名称： 基于框架提升变换的多源图像融合研究

项目编号： No.61501029

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 石岩

作者单位： 北京理工大学

项目金额： 19万元

中文摘要： 随着现代社会信息化进程的加快及多传感器应用平台的普及，信息呈现海量、多样、冗余等特点。如何充分高效地利用多源信息已成为当前亟需解决的问题。在此背景下，多源图像融合技术已日渐突显出其重要性，并引发了众多科研工作者广泛深入地研究。目前在学术界，基于多分辨分析的融合方法最受关注，其经历了从小波变换到多尺度几何变换的发展过程。并已证实，多尺度、方向性、冗余性、平移不变性以及局部自适应性等是有利于融合的重要因素。然而目前，已有的多尺度变换尚存在一些不足，这促使本项目研究一类新颖的变换，即小波框架提升变换。本项目研究小波框架提升构造原理，在此基础上构造各向同性、各向异性及分数阶框架提升变换，以期获得图像多尺度、多方向、多阶次等不同层面的时频信息表征，为图像融合创造有利条件。进一步，研究基于框架提升变换域的融合算法，探索具有一般性、普适性的多源融合机制，为多源图像融合提供新思路。

中文关键词： 框架提升变换；图像融合；多传感器；稀疏表示；分数域

英文摘要： With the acceleration of informatization of the modern society and the popularity of multi-sensor application platforms, information features massiveness, diversity and redundancy. How to make efficient use of multiple sources of information has become an urgent problem to be solved. Against this backdrop, the importance of multi-source image fusion technique has been increasingly highlighted, which led a large number of researchers to study extensively. Currently, most attention in academia has been paid to the multiresolution-based image fusion, which has been experiencing a development from wavelet transform to multi-scale geometric transforms. And it has been proved that several characteristics such as multi-scale, directionality, redundancy, translation invariance, and local adaptability are the key factors that are beneficial for fusion. However there still exist some issues among the various existing transforms. This project is prompted to study a novel class of transforms called framelet lifting transforms. The project investigates the principle of lifting construction of wavelet frames as well as three types of framelet lifting transforms which inhere isotropic, anisotropic, and fractional characteristics respectively, aiming at providing time-frequency information in different aspects including multi-scale, multi-direction and multi-order, and providing beneficial conditions for image fusion. Furthermore, the project studies the image fusion methods based on framelet lifting transform, aims to explore the general, universal fusion mechanism, and to provide new ideas for multi-source image fusion.

英文关键词： framelet lifting transform;image fusion;multi-sensor;sparse representation;fractional domain