Convolution sparse coding induced prior for hyperspectral image super resolution

doi:10.1016/j.jvcir.2026.104939

Convolution sparse coding induced prior for hyperspectral image super resolution

2026-09-01

0

PRE

AI

D

Dibyajyoti Guha *

DOI：10.1016/j.jvcir.2026.104939

Original

Original request for help

Save

Abstract

En 中文

In this paper, we propose an efficient method for hyperspectral image super resolution from a low spatial resolution hyperspectral image and high spatial resolution RGB image. This paper considers a convolutional sparse coding induced prior and guided filter aided super resolved image as prior for generating super resolution image from LR HSI and HR MSI. Most of the works have focused on mapping the sparse codes between LR HSI and HR MSI. We explore a new way of retaining the smooth components and residual components of the LR HSI using guided filter based approach. Our work utilizes convolution sparse coding guided proximal term to enforce structural sparsity to estimate the abundance matrix. Guided filter based super resolved image is used as a prior term. We are able to improve PSNR in the range of 0.1-0.15 while comparing with state-of-the-art methods for CAVE and Harvard dataset.

Keywords:

Hyperspectral image super resolution

Convolutional sparse coding

Non negative matrix factorization

Alternating direction method of multipliers

Sparse coding

Dictionary learning

Convolution sparse coding induced prior for hyperspectral image super resolution

Abstract

Journal

Organization

Cited Papers