Physically Plausible Spectral Reconstruction from RGB Images
This work solves the issue of physically implausible spectral reconstruction for computer vision and imaging applications, representing an incremental improvement by enforcing physical constraints and handling exposure variations.
The paper tackled the problem of spectral reconstruction from RGB images by addressing the lack of physical plausibility and sensitivity to exposure changes in existing CNN-based methods, achieving state-of-the-art recovery performance with exact reintegration to input RGBs under varying exposures.
Recently Convolutional Neural Networks (CNN) have been used to reconstruct hyperspectral information from RGB images. Moreover, this spectral reconstruction problem (SR) can often be solved with good (low) error. However, these methods are not physically plausible: that is when the recovered spectra are reintegrated with the underlying camera sensitivities, the resulting predicted RGB is not the same as the actual RGB, and sometimes this discrepancy can be large. The problem is further compounded by exposure change. Indeed, most learning-based SR models train for a fixed exposure setting and we show that this can result in poor performance when exposure varies. In this paper we show how CNN learning can be extended so that physical plausibility is enforced and the problem resulting from changing exposures is mitigated. Our SR solution improves the state-of-the-art spectral recovery performance under varying exposure conditions while simultaneously ensuring physical plausibility (the recovered spectra reintegrate to the input RGBs exactly).