基于PCA和总变差模型的图像融合框架
An Image Fusion Framework Based on Principal Component Analysis and Total Variation Model
-
摘要: 为了能够抑制融合图像中的噪声且提升融合效果,提出一种在进行融合时去噪的新框架,并在此基础上提出一种基于主成分分析(PCA)的融合框架.首先将源图像进行PCA操作,依据前几个主成分重建图像,再经下采样过程得到近似图像;然后通过上采样得到与上层图像的差异图像,即细节图像;最后将最底层近似图像与各层细节图像累加,完成图像的重构.将该框架纳入总变差模型后形成一种新的具有融合和去噪效果的框架.实验结果表明,该方法不仅能对同分辨率图像融合,获得较好的融合效果,而且在全色图像和多光谱图像的融合中可较好地保持光谱信息和空间信息,并能够抑制图像中存在的噪声.Abstract: To resist the noises and improve the quality of the fusion image,a new framework of simultaneous denoising during the image fusion is proposed.Based on the denoising framework,an image fusion framework based on Principal Component Analysis(PCA) is adopted.PCA is first applied to the source images.Then approximate images are acquired after subsampling the image reconstructed with the first-n components;the difference between current and the previous level approximate images is computed to construct a detail image.In the end,the last approximate image and all detail images are merged in form a new fused image.Embedding this fusion framework in the Total Variation model,a new framework that can fuse and denoise images simultaneously is formed.The experimental results demonstrate that this framework not only has good performance in image fusion of the same resolution,but also retain the spectral information and space information in the fusion of panchromatic and multi-spectral images.Additionally,it can restrain noises in the source images.