对偶框架各向异性提升变换理论与应用研究

项目名称： 对偶框架各向异性提升变换理论与应用研究

项目编号： No.61271010

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 无线电电子学、电信技术

项目作者： 杨小远

作者单位： 北京航空航天大学

项目金额： 60万元

中文摘要： 框架变换是继小波变换之后调和分析重要的研究领域,目前国内外学者还没有开展对偶框架各向异性提升变换的研究.提升变换一个重要特征是变换的灵活性,可以自适应表示图像不同子区域的局部特征,为此本项目将研究其提升变换理论,在此基础上研究对偶框架的各向异性提升变换，构造具有不同阶的局部预测和更新算子以及对偶框架各向异性提升变换的基函数,研究其变换性能.提出对偶框架提升和各向异性提升变换的理论框架,本项目理论研究的实际意义是有效地表示遥感图像的多方向信息和轮廓特征,实现遥感图像的稀疏表示.在理论研究基础上,通过研究基于遥感图像正则性的自适应图像分割和方向插值加权技术,构造面向遥感图像几何特征的对偶框架各向异性提升变换基函数，提出基于对偶框架各向异性提升变换多源遥感图像(SAR与可见光图像、SAR图像与多光谱图像、多波段SAR图像的融合)融合算法.本项目理论研究成果可以广泛应用到遥感图像处理领域.

中文关键词： 对偶框架；提升变换；各向异性多尺度变换；图像去噪；多源遥感图像融合

英文摘要： The wavelet frame transform is an important area of research of harmonic analysis following the wavelet transform. The lifting scheme of wavelet bi-frame has not been investigated yet so far both here and abroad. One of the important features of the lifting scheme is that the flexibility and the ability to characterize the local features of images. This project aims to study the lifting scheme of wavelet bi-frame as well as the anisotropic property of wavelets. Several local prediction and update operators with different orders as well as the basis functions of the anisotropic bi-frame lifting transform are constructed. As a consequence, a theoretical framework of the wavelet bi-frame and anisotropic lifting transform is proposed. The practical significance of the theoretical study of this project is to realize the sparse representation of remote sensing images that contain rich orientations and contours more effectively. Based on the theoretical study, by studying the adaptive image segmentation and directional weighted interpolation techniques based on the regularity of remote sensing images, the anisotropic basis function of bi-frame lifting transform facing remote sensing images that contain geometric features is constructed. On the other hand, the image fusion strategies of remote sensing images based on t

英文关键词： bi-frame；lifting transform；anisotropic multi-scale transform；image denoising；multi-source remote sensing image fusion