MST++: Multi-stage Spectral-wise Transformer for Efficient Spectral Reconstruction

Existing leading methods for spectral reconstruction (SR) focus on designing deeper or wider convolutional neural networks (CNNs) to learn the end-to-end mapping from the RGB image to its hyperspectral image (HSI). These CNN-based methods achieve impressive restoration performance while showing limitations in capturing the long-range dependencies and self-similarity prior. To cope with this problem, we propose a novel Transformer-based method, Multi-stage Spectral-wise Transformer (MST++), for efficient spectral reconstruction. In particular, we employ Spectral-wise Multi-head Self-attention (S-MSA) that is based on the HSI spatially sparse while spectrally self-similar nature to compose the basic unit, Spectral-wise Attention Block (SAB). Then SABs build up Single-stage Spectral-wise Transformer (SST) that exploits a U-shaped structure to extract multi-resolution contextual information. Finally, our MST++, cascaded by several SSTs, progressively improves the reconstruction quality from coarse to fine. Comprehensive experiments show that our MST++ significantly outperforms other state-of-the-art methods. In the NTIRE 2022 Spectral Reconstruction Challenge, our approach won the First place. Code and pre-trained models are publicly available at https://github.com/caiyuanhao1998/MST-plus-plus.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Spectral Reconstruction ARAD-1K MST++ PSNR 34.32 # 1
MRAE 0.1645 # 1
RMSE 0.0248 # 1
Spectral Reconstruction CAVE MST++ PSNR 35.99 # 3
SSIM 0.951 # 3
Spectral Reconstruction KAIST MST++ PSNR 35.99 # 3
SSIM 0.951 # 3
Spectral Reconstruction Real HSI MST++ User Study Score 13 # 3

Methods