Abstract: Computational color constancy refers to the estimation of the scene illumination and makes the perceived color relatively stable under varying illumination. In the past few years, deep Convolutional Neural Networks (CNNs) have delivered superior performance in illuminant estimation. Several representative methods formulate it as a multi-label prediction problem by learning the local appearance of image patches using CNNs. However, these approaches inevitably make incorrect estimations for the ambiguous patches affected by their neighborhood contexts. Inaccurate local estimates are likely to bring in degraded performance when combining into a global prediction. To address the above issues, we propose a contextual deep network for patch-based illuminant estimation equipped with refinement. First, the contextual net with a center-surround architecture extracts local contextual features from image patches, and generates initial illuminant estimates and the corresponding color corrected patches. The patches are sampled based on the observation that pixels with large color differences describe the illumination well. Then, the refinement net integrates the input patches with the corrected patches in conjunction with the use of intermediate features to improve the performance. To train such a network with numerous parameters, we propose a stage-wise training strategy, in which the features and the predicted illuminant from previous stages are provided to the next learning stage with more finer estimates recovered. Experiments show that our approach obtains competitive performance on two illuminant estimation benchmarks.

Sensor-Independent Illumination Estimation for DNN Models

Automatical Enhancement and Denoising of Extremely Low-light Images.

NeIF: Representing General Reflectance as Neural Intrinsics Fields for Uncalibrated Photometric Stereo

Self-adaptive Single and Multi-illuminant Estimation Framework based on Deep Learning

Effective Cross-Sensor Color Constancy Using a Dual-Mapping Strategy.

DeepIlluminance: Contextual Illuminance Estimation via Deep Neural Networks.

Deep Separable Convolution Neural Network For Illumination Estimation

Practical cross-sensor color constancy using a dual-mapping strategy

Multiple Illumination Estimation With End-To-End Network

Learning to predict indoor illumination from a single image

Learning Enriched Illuminants for Cross and Single Sensor Color Constancy

Single and Multiple Illuminant Estimation Using Convolutional Neural Networks

Dive into Illuminant Estimation from a Pure Color View

Object-Based Illumination Estimation with Rendering-Aware Neural Networks

Reconstruction of spectral irradiance in a real application with a multi-channel spectral sensor using convolutional neural networks.

A study of the effect of the illumination model on the generation of synthetic training datasets

Biologically inspired image invariance guided illuminant estimation using shallow and deep models

Self-Supervised Low-Light Image Enhancement Using Discrepant Untrained Network Priors

Multi Illumination Color Constancy Based on Multi-Scale Supervision and Single-Scale Estimation Cascade Convolution Neural Network.

Image Segmentation for Night-Vision Surveillance Camera Based on Deep Learning

Learning deep illumination-robust features from multispectral filter array images