Color Correction for Multi-view Video Based on Background Segmentation and Dominant Color Extraction

Color correction is a necessary operation in multi-view video processing, because color tends to be influenced by camera characteristic, surface reflectance or scene illumination. To achieve high-quality correction results, a new color correction method, based on the theoretical model of multi-view imaging and image restoration, is proposed in this paper. Considering the illumination problem in multi-view imaging, foreground and background regions are separated from images and dominant color extraction is used only for background regions of reference and input images, so that uniform reference surface information is used. Then, color correction is extended to video sequences with a tracking approach. Furthermore, an objective performance evaluation is proposed to evaluate the color correction. We present a variety of results for different test sequences, arguing that background-based method outperforms foreground-based method, and better subjective and objective visual effect can be achieved for images as well as videos. Key-Words: Multi-view video, color correction, background segmentation, dominant color extraction, principal component analysis, color difference