The Research On Color Correction Of Multi-view Video Coding

As the rapid development and wide application of digital video technology, traditional two-dimensional plane video can not meet the basely requirements, such as true and natural reproduction of the scene, so 3D stereoscopic display technology came into being. As the key technology of 3D stereoscopic display, multi-view video have recently received increasing attention from both academic and industry field. In multi-view video system, multiple video cameras are used to capture the same scene, different from traditional one-view video. Efficient compression methods should be used, because of the huge amount of data captured in multi-view video system. But there may be some color variation of images among different views as the impact of some conditions, for example, the different parameters of different cameras, light environment and so on, so it's hard to search the exactly matching points between the pictures from different views when encoding.The importance of color correction in multi-view video coding is firstly introduced in the paper, and then several algorithms on color correction are proposed. The first one is global linear correction based on matching points, it corrects the sequences, which have significant global color variations between views. The second one is color cast correction, firstly, we search the matching points between two images from different views, and test the color cast phenomena based on Von Kries modal, and then do color correction. The third one is histogram matching based on RGB color space, in this algorithm we first transform the sequences from YUV color space to RGB color space, and then do histogram matching for the three channel R,G,B separately. The forth one is do color correction based on matching points and least squares method, the chiefly step is search the matching points among images of all views, calculate the average of these points, and then correct all pictures use the coefficients which are calculated based on the method of least squares. At last two improved algorithm are proposed based on the analyses of existed color correction algorithm. The first one is histogram matching based on RGB color space with matching points, in this algorithm we only use the matching points, searched from block matching between two images, to calculate the histogram, instead of all the points in this image. The second one is do color correction based on adaptive disparity vector and least square method, which is an improvement for color correction based on matching points and least square method. In this algorithm different search ranges for different sequences and views are firstly determined adaptively, so as to improve the adaptability and achieve relative balance between objective quality and subjective quality. Detail theory proof, experiment images and data are provided for all of the algorithms in the paper.Finally, to intuitively observe the difference of color between images, a software based on VC/MFC is developed. With it, we can open and play several videos at the same time, show the Y,U,V values of each pixels separately. Further more, difference image for two images can be produced, and contrast display pixel information mean while.