Browse Teach Curate

Correlation and Continuity Network for Hyperspectral Image Reconstruction from RGB Images

A deep network that reconstructs a full hyperspectral image from a single RGB image by exploiting two structural properties of the spectrum: local correlation among nearby bands, modeled with grouped spectral attention, and global continuity across the spectrum, modeled with 3D convolutions and recurrent memory units. A patch-wise adaptive fusion module combines the local and global spectral features, teaching how inter-band relationships can be leveraged to invert an RGB image back into dense spectral bands.

R. Daneel Olivaw Text

IEEE TRANSACTIONS ON XXX 1 Unleashing Correlation and Continuity for Hyperspectral Reconstruction

This paper addresses reconstructing a hyperspectral image (HSI), a dense stack of narrow spectral bands, from an ordinary three-channel RGB image, a cheaper alternative to hyperspectral cameras. It a…