Colour harmonisation for images and videos via two-level graph cut
Zhen Tang, Zhiwei Miao, Yanli Wan, Forrest Fabian Jesse · IET Image Processing · 2011
A two-level graph cut algorithm is presented to deal with the coherence preservation problem that appears when trying to perform colour harmonisation on images and videos. Region-level graph cut is a technique used to avoid harmonising disconnected areas belonging to the same object with different harmonic hues. Pixel-level graph cut is a technique used to obtain a spatially coherent harmonisation. The authors extend this algorithm to videos by employing optical flow, region growing and foreground extraction algorithms, and present a temporally consistent video harmonisation method. The proposed approach overcomes several shortcomings of the existing methods, and the experiments demonstrate the effectiveness of it.