Deep spectral reflectance and illuminant estimation from self-interreflections

Abstract

In this work, we propose a CNN-based approach to estimate the spectral reflectance of a surface and the spectral power distribution of the light from a single RGB image of a V-shaped surface. Interreflections happening in a concave surface lead to gradients of RGB values over its area. These gradients carry a lot of information concerning the physical properties of the surface and the illuminant. Our network is trained with only simulated data constructed using a physics-based interreflection model. Coupling interreflection effects with deep learning helps to retrieve the spectral reflectance under an unknown light and to estimate the spectral power distribution of this light as well. In addition, it is more robust to the presence of image noise than the classical approaches. Our results show that the proposed approach outperforms the state of the art learning-based approaches on simulated data. In addition, it gives better results on real data compared to other interreflection-based approaches.

Triple-layer remote phosphor structure: a selection of the higher color quality and lumen efficiency for WLEDs

“…Therefore, to confirm the benefits of the TL to the luminous flux, we carried out a mathematic system of transmitted blue light and converted yellow light in the triple-layer structure and presented below as the following. The transmitted blue light and converted yellow light in dual-layer structure with each phosphor layer having the thickness of h, the calculation is expressed as [22], [23]: ℎ, these mentioned blue light and yellow light can be computed by:…”

Real‐time skin chromophore estimation from hyperspectral images using a neural network

“…In our approach, the neural network was not trained to model light‐skin interactions, but only to solve an inversion problem more quickly than by using optimization‐based algorithm. The approach illustrated in Figure 1A, using a training dataset based on a Monte Carlo method 7‐10,24 and a much more detailed model of skin than our two‐layer model, could be investigated. In this case, the neural network would provide a model for skin spectral reflectance and solve the inverse problem as well.…”

Interreflections

“…Alternatively, one can constrain the problem by using fluorescent paint and imaging the surface under different colored light sources [12]. More recent methods have attempted to learn a mapping from the image to the spectral reflectance and illuminant by training on many synthetic examples using the full infinite bounce model [13]. For further examples of interreflections in spectral estimation, see the following review [14].…”