HDMba: Hyperspectral Remote Sensing Imagery Dehazing with State Space Model

Hang Fu,Genyun Sun,Yinhe Li,Jinchang Ren,Aizhu Zhang,Cheng Jing,Pedram Ghamisi

2024-06-09

Abstract:Haze contamination in hyperspectral remote sensing images (HSI) can lead to spatial visibility degradation and spectral distortion. Haze in HSI exhibits spatial irregularity and inhomogeneous spectral distribution, with few dehazing networks available. Current CNN and Transformer-based dehazing methods fail to balance global scene recovery, local detail retention, and computational efficiency. Inspired by the ability of Mamba to model long-range dependencies with linear complexity, we explore its potential for HSI dehazing and propose the first HSI Dehazing Mamba (HDMba) network. Specifically, we design a novel window selective scan module (WSSM) that captures local dependencies within windows and global correlations between windows by partitioning them. This approach improves the ability of conventional Mamba in local feature extraction. By modeling the local and global spectral-spatial information flow, we achieve a comprehensive analysis of hazy regions. The DehazeMamba layer (DML), constructed by WSSM, and residual DehazeMamba (RDM) blocks, composed of DMLs, are the core components of the HDMba framework. These components effectively characterize the complex distribution of haze in HSIs, aiding in scene reconstruction and dehazing. Experimental results on the Gaofen-5 HSI dataset demonstrate that HDMba outperforms other state-of-the-art methods in dehazing performance. The code will be available at <a class="link-external link-https" href="https://github.com/RsAI-lab/HDMba" rel="external noopener nofollow">this https URL</a>.

Computer Vision and Pattern Recognition,Image and Video Processing

What problem does this paper attempt to address?

The problem addressed in this paper is the removal of haze in hyperspectral remote sensing images (HSI). Haze in HSI can cause decreased spatial visibility and spectral distortion, and existing haze removal methods often struggle to balance global scene restoration, local detail preservation, and computational efficiency. The paper proposes a new network called HDMba, which is the first framework to utilize the Mamba model for HSI haze removal. HDMba designs a Window Selective Scanning Module (WSSM) to capture local dependencies within the window and global correlations between windows, improving the capability of conventional Mamba in local feature extraction. By modeling local and global spectral-spatial information flow, HDMba is able to comprehensively analyze haze regions and achieve effective haze removal. Experimental results demonstrate that HDMba outperforms other state-of-the-art methods in haze removal performance for hyperspectral images.