Abstract:How to improve the spatial resolution as much as possible while maintaining the spectral information of multi-spectral (MS) image in the field of image fusion is of great significance for practical applications, such as map updating, feature classification, and target recognition. To analyze the coefficients of the subband distribution characteristics, in this paper, we propose a new panchromatic (PAN) and MS image pansharpening model based on an adaptive neural network and sparse representation in the non-subsample shearlet transform (NSST) domain. First, this algorithm is specific to regional directional characteristics in the high-frequency subband of PAN and MS images, and we propose an adaptive pulse coupled neural network (PCNN) model. The model can adaptively calculate the link strength of a neural cell based on the region energy. Furthermore, we apply the model to the high-frequency fusing process with the corresponding fusion rule, and the rule can distinguish the high-frequency coefficients by ignition times, which can more effectively capture the geometric texture information and detailed information in the PAN image, enhancing the spatial resolution of the fused image. Second, because of the low-frequency sub-bands from the PAN image and I component obtained by intensity-hue-saturation (IHS) transformation of the MS images with high similarity to the original image but poor sparsity, we select a set of PANimages for learning, a more targeted over-complete dictionary for low-frequency sub-band sparse representation is obtained. Then, the larger absolute value of the sparse matrix is selected to obtain the low-frequency coefficients for the fusion image while maintaining the MS spectral information effectively, and the representation of characteristic information of low-frequency subband is more effective. A large number of simulation experiments verify the effectiveness of the proposed method.

The PAN and MS Image Pansharpening Algorithm Based on Adaptive Neural Network and Sparse Representation in the NSST Domain

Efficient and Accurate Hyperspectral Pansharpening Using 3D VolumeNet and 2.5D Texture Transfer

SSETPAN: Spatial-Spectral Enhanced Transformer Based Network for Pansharpening

Detail Information Prior Net for Remote Sensing Image Pansharpening

A Fusion Method of Multi-spectral Image and Panchromatic Image Based on NSCT and IHS Transform

Multispectral And Panchromatic Images Fusion Based On Nsct And Gs Transform

A new technique for multispectral and Panchromatic image fusion

Pan-sharpening via regional division and NSST

Multispectral and Panchromatic Image Fusion Based on Unsubsampled Contourlet Transform

Pansharpening with a Gradient Domain GIF Based on NSST

An Efficient Pansharpening Approach Based on Spatial Information and Multi-Layer Fusion.

Pansharpening via Frequency-Aware Fusion Network with Explicit Similarity Constraints

Fusion of Panchromatic and Multispectral Images via Morphological Operator and Improved PCNN in Mixed Multiscale Domain

Multiscale Spatial-Frequency Domain Dynamic Pansharpening of Remote Sensing Images Integrated With Wavelet Transform

PWNet: an Adaptive Weight Network for the Fusion of Panchromatic and Multispectral Images.

Asymmetric Bidirectional Fusion Network for Remote Sensing Pansharpening

Pansharpening Via Triplet Attention Network with Information Interaction

A pansharpening method combining iterative filtering and NSST-NSML-PAPCNN to optimize spatial detail extraction and injection

Multispectral Pansharpening Based on Multisequence Convolutional Recurrent Neural Network

Progressive High-Frequency Reconstruction for Pan-Sharpening with Implicit Neural Representation