Remote Sensing Image Fusion Based on Non-Negative Matrix Decomposition and Non-Subsampled Shear Wave Transform

This article proposes an image fusion method based on non-negative matrix factorization (NMF) and non-subsampled shearlet transform (NSST). The method combines NMF and NSST to extract the common and structural information of images, resulting in clearer, more natural, and accurate image fusion results. Specifically, we use NMF to decompose multiple input images and obtain their common and individual parts. Through the decomposition of NMF, the original image can be represented as a product of a non-negative matrix V , which contains the pixel values of the original image and several columns representing the basis matrix of the image. In our method, the common part corresponds to the information shared by the images, while the individual part corresponds to the individual features of the images. Then, we use NSST to decompose the common and individual parts to obtain image information at different scales and orientations. NSST is a multiscale analysis method based on shearlet transform, which can preserve the structural information of the image and suppress the pseudo-Gibbs phenomenon. Finally, we fuse the common and individual parts processed by NSST separately to obtain an output image that integrates the details and features of the image. This fusion method can handle images of different types, sizes, and resolutions well and performs well in processing complex situations. To verify the performance of the proposed method, we conducted experiments on different datasets and compared them with other commonly used image fusion methods. The experimental results show that the proposed method achieves good fusion effects. Therefore, the method proposed in this article has a wide range of application prospects in the field of image fusion.