Effective Learning-Based Illuminant Estimation Using Simple Features


Dongliang Cheng       Brian Price       Scott Cohen       Michael S. Brown

Abstract:

Illumination estimation is the process of determining the chromaticity of the illumination in an imaged scene in order to remove undesirable color casts through white-balancing. While computational color constancy is a well-studied topic in computer vision, it remains challenging due to the ill-posed nature of the problem. One class of techniques relies on low-level statistical information in the image color distribution and works under various assumptions (e.g. Grey-World, White-Patch, etc). These methods have an advantage that they are simple and fast, but often do not perform well. More recent state-of-the-art methods employ learning-based techniques that produce better results, but often rely on complex features and have long evaluation and training times. In this paper, we present a learning-based method based on four simple color features and show how to use this with an ensemble of regression trees to estimate the illumination. We demonstrate that our approach is not only faster than existing learning-based methods in terms of both evaluation and training time, but also gives the best results reported to date on modern color constancy data sets.


Effectiveness and efficiency:

Evaluation time vs. performance of representative illuminant estimation methods. Statistics-based methods are fast but have lower accuracy than learning-based methods. The slow speed of learning-based methods makes them impractical for onboard camera white-balancing. Our proposed learning-based method achieves high accuracy and fast evaluation. (Mean angular error and time statistics for this plot are based results in Table 1 and Table 3 in the paper. For more details, please refer to the paper). Note that the time axis is nonlinear.


Publication:

Effective Learning-Based Illuminant Estimation Using Simple Features (PDF, Supplementary), CVPR, 2015.


Source Code and Results:

Click here to download the matlab source code of the proposed algorithm.

Click here to download the experiment results in the paper.


Acknowledgements:

This work was supported by Singapore A*STAR PSF grant 11212100 and an Adobe gift award.


This page was last modified on: May 11, 2015