Pansharpening is the process of fusing a low-resolution multispectral (LRMS) image with a high-resolution panchromatic (PAN) image. In the process of pansharpening, the LRMS image is often directly upsampled by a scale of 4, which may result in the loss of high-frequency details in the fused high-resolution multispectral (HRMS) image. To solve this problem, we put forward a novel progressive cascade deep residual network (PCDRN) with two residual subnetworks for pansharpening. The network adjusts the size of an MS image to the size of a PAN image twice and gradually fuses the LRMS image with the PAN image in a coarse-to-fine manner. To prevent an overly-smooth phenomenon and achieve high-quality fusion results, a multitask loss function is defined to train our network. Furthermore, to eliminate checkerboard artifacts in the fusion results, we employ a resize-convolution approach instead of transposed convolution for upsampling LRMS images. Experimental results on the Pléiades and WorldView-3 datasets prove that PCDRN exhibits superior performance compared to other popular pansharpening methods in terms of quantitative and visual assessments.

Pansharpening is the process of fusing low spatial resolution multispectral (MS) images with high spatial resolution panchromatic (PAN) images, so as to obtain high spatial resolution multispectral (HRMS) images. In this study, a new pansharpening method based on a joint-guided detail extraction is proposed to maintain the spectral and spatial fidelity of a pansharpened image. First, to obtain details that are highly correlated with an MS image, a new PAN image is constructed and guided by the intensity component of the MS image through a variational model. The construction of the new PAN image improves the correlation between the PAN and MS images, and thus reduces the spectral distortion. The variational model is rapidly solved using the least-squares method. Second, to obtain accurate details from the new PAN image, the extraction of the details is guided by each band of the MS image through a regression model, which can further reduce the spatial distortion. The regression model is effectively solved using the gradient descend method. Finally, the details are injected into the up-sampled MS image to obtain a fused image. Numerous experiments on the proposed approach were conducted and the results were compared with previous state-of-the-art pansharpening methods. The experimental results verify that the proposed method can efficiently achieve high-quality HRMS images.

Pansharpening exploits the high-frequency component (HFC) of panchromatic (PAN) images to restore the spatialresolution of the corresponding multispectral (MS) image. In this article, a dual-injection model-based multiband remote sensing image pansharpening method is presented that focuses on how to correctly use the HFC to improve the MS image for obtaining a high-spatial resolution and MS image. The model is based on a two-step HFC injection algorithm with two different injection gains. In the first step, an HFC is reconstructed with sparse theory, and an injection gain based on the relationship between PAN and MS images is developed. Employing the previous injection gain and the reconstructed HFC on an upsampling MS image, an improved LRMS (ILRMS) image is then produced. In the second step, another injection gain based on the differences and similarities between the PAN and MS images is designed. With the help of this injection gain, the fusion image is achieved via the adaptive integration of the ILRMS image and the HFC from the PAN image. Experiments confirm that the proposed method is more effective than some popular widely used pansharpening methods.

No abstract