一种基于多尺度稀疏分解的遥感图像融合新方法

发布时间：2018-04-03 15:38

本文选题：多尺度　切入点：稀疏分解　出处：《国土资源遥感》2017年03期

【摘要】：为取得更好的遥感图像融合效果,结合形态成分分析的思想,提出了图像的多尺度稀疏分解方法。集合曲波变换基和局部离散余弦变换基组成分解字典,通过控制字典系数的大小,将二维图像从多个尺度稀疏分解为纹理成分和卡通成分;从图像融合的信息量角度出发,提出了基于多尺度稀疏分解的遥感图像融合方法,通过稀疏分解提取有效尺度下高空间分辨率图像纹理成分和多光谱图像卡通成分,并对二者进行稀疏重建得到融合图像。与已有的经典融合方法相比,该方法以较小的计算代价换取了更高的空间分辨率和更低的光谱失真;与稀疏重建法相比,该方法的执行速率有较大提升,且有更好的融合效果。因此,所提出的基于多尺度稀疏分解的遥感图像融合方法有一定的推广应用价值。
[Abstract]:In order to obtain better fusion effect of remote sensing images, a multi-scale sparse decomposition method is proposed based on morphological component analysis.Set Qu Bo transform basis and local discrete cosine transform basis to form decomposition dictionary. By controlling the size of dictionary coefficient, two-dimensional image is decomposed from multiple scales sparse to texture and cartoon components.A method of remote sensing image fusion based on multi-scale sparse decomposition is proposed. Texture components of high-resolution images and cartoon components of multi-spectral images are extracted by sparse decomposition at effective scales and the fusion images are obtained by sparse reconstruction of the two images.Compared with the classical fusion methods, the proposed method gains higher spatial resolution and lower spectral distortion at a lower computational cost, and improves the speed and fusion performance of the proposed method compared with the sparse reconstruction method.Therefore, the proposed remote sensing image fusion method based on multi-scale sparse decomposition has a certain value of popularization and application.
【作者单位】：烟台大学计算机与控制工程学院;
【基金】：山东省自然科学基金项目“基于稀疏盲图像分离的多源遥感影像融合”(编号:ZR2014FQ026)资助
【分类号】：TP751

【相似文献】