Training-Based Spectral Reconstruction from a Single RGB Image

Abstract

This paper focuses on a training-based method to reconstruct a scene's spectral reflectance from a single RGB image captured by a camera with known spectral response. In particular, we explore a new strategy to use training images to model the mapping between camera-specific RGB values and scene reflectance spectra. Our method is based on a radial basis function network that leverages RGB white-balancing to normalize the scene illumination to recover the scene reflectance. We show that our method provides the best result against three state-of-art methods, especially when the tested illumination is not included in the training stage. In addition, we also show an effective approach to recover the spectral illumination from the reconstructed spectral reflectance and RGB image. As a part of this work, we present a newly captured, publicly available, data set of hyperspectral images that are useful for addressing problems pertaining to spectral imaging, analysis and processing.

Publications:


Dataset

Please download dataset at here.

This dataset contains 66 hyperspectral images (in *.mat format) which were used for training and testing in our paper. Each hyperspectral image contains the spectral irradiance and spectral illumination (400-700 nm with step of 10 nm). For more information, please refer to the ReadMe.txt file.


Matlab Code

The new version of Matlab code has been updated here to fix the resolution error. Please make sure your computer have enough RAM (at least 16GB) and Matlab 2013a to run this software.


People

Rang NGUYENnguyenho at comp.nus.edu.sg
Dilip K. PRASADdilipprasad at gmail.com
Michael S. BROWNbrown at comp.nus.edu.sg

Last updated: 21 August 2014