A colour correction preprocessing method for multiview video coding

In multiview video, a number of cameras capture the same scene from different viewpoints. There can be significant variations in the colour of views captured with different cameras, which negatively affects performance when the videos are compressed with inter-view prediction. In this paper, a method is proposed for correcting the colour of multiview video sets as a preprocessing step to compression. The corrected YUV values of a pixel are expressed as a weighted sum of the original YUV values. Disparity estimation is used to find matching points between a reference view and the view being corrected. A least squares regression is performed on these sets of matching points to find the optimal weight parameters that will make the current view most closely match the reference. Experimental results show that the proposed method produces colours that closely match the reference view. Furthermore, when multiview video is compressed with H.264 using inter-view prediction, the proposed method increases compression efficiency by up to 1.0 dB compared to compressing the original uncorrected video.