Dual-Illumination Weighting and Estimation | IEEE Conference Publication | IEEE Xplore

Dual-Illumination Weighting and Estimation


Abstract:

Illumination estimation refers to estimating the chromaticity vector of illumination, and can be used to recover the surface color under white light. Dual-illuminant is a...Show More

Abstract:

Illumination estimation refers to estimating the chromaticity vector of illumination, and can be used to recover the surface color under white light. Dual-illuminant is a common scenario in computational illumination estimation tasks. A straightforward way to correct the dual-illuminant image can be estimating a spatially-varying illumination map. However, it is hindered by the lack of large-scale annotated datasets for data-driven methods. In this paper, we propose a novel approach to obtain the dominant dual-illuminant and the pixel-wise illuminant map on real dual-illuminant raw images. Our method consists of 1) dual-illuminant image generator (DIG) to synthesize dual-illuminant images from unique-illuminant datasets assuming the Lambertian model; 2) dual-illuminant estimation network (DE-Net) to estimate illuminant both globally and locally. Quantitative experiments show that with DIG synthesized dual-illuminant images, DE-Net obtains the best accuracy in dual-illumination detection and estimation on the Gehalr-shi dataset and Mutlti-Illuminant Multi-Object dataset.
Date of Conference: 21-25 August 2022
Date Added to IEEE Xplore: 29 November 2022
ISBN Information:

ISSN Information:

Conference Location: Montreal, QC, Canada

References

References is not available for this document.