Illuminant classification based on random forest
Published in 14th IAPR International Conference on Machine Vision Applications (MVA), 2015
We present a novel machine learning/pattern recognition based colour constancy method. We cast colour constancy as an illumination source recognition problem, and have developed an effective and efficient random forest based classification technique for inferring the class of illumination source of an image. In an opponent colour space, we have developed a binary image representation feature that is somewhat insensitive to image contents for building the random forest classifier that infers the likely class of the illumination source of the image. The binary image feature and the tree structure of the recognition system are intrinsically efficient. We present results on colour constancy benchmark data sets and show that our new technique outperforms state of the art techniques.
Recommended citation: Liu, Bozhi, and Guoping Qiu. “Illuminant classification based on random forest.” 2015 14th IAPR International Conference on Machine Vision Applications (MVA). IEEE, 2015.