查看文章

In this paper, we propose the grey-wavelet hypothesis for colour constancy which assumes that both the average reflectance of surfaces and the average of reflectance differences in a scene are achromatic. The framework combines well-known grey-world and grey-edge hypotheses under the wavelet framework. The quality of the various instantiations of the framework is evaluated on set of images of colourful objects under a controlled indoor setting and on real-world datasets containing mainly outdoor scenes. The experiments show that the proposed colour constancy algorithm obtains better results than the state-of-the-art colour constancy algorithms of similar complexity. Since the framework operates in wavelet transform domain, it can be easily adopted to JPEG2000 for the purpose of colour constancy.