In order to solve non-consistence color appearance between different viewpoint images, a multi-view video automatic color correction system is proposed. Global correction matrix is first obtained between target image and source image, then estimate whether global correction is satisfied. If no global correction is satisfied, image segmentation and Karhunen-Loeve transform are performed. Then local mapping relations between target image and source image segmentation regions can be established. And preferred region color mapping is performed to correct source image. Finally video tracking technique is used to correct video images. For the standard multi-view images, compared with histogram matching and global one dimension correction algorithms, experimental results show that the proposed algorithm can effectively eliminate the influence of matching distortion, and achieve good correction result. The studies indicate that the system can apropos reveal color change relation existing in images, has capability of showing content adaptivity, and also is important method of color correction.