Research On Color Correction Of Multi - View Video

Nowadays, multi-view video develop rapidly and bring us the vivid reflection of the scene. Multiple cameras are located at different angles for capturing the information of the same scene, thus massive data of information is produced. In the process of video capturing, due to the difference of the reflective on the surface of the object, noise environment and imaging characteristics of the device itself(including the charge-coupled device CCD noise, jitter, shutter speed, and exposure time, etc.), lead to the color inconsistency between the various viewpoints of multi-view video. Thus, it is unable to remove the redundant information between the various viewpoints and drawing a direct impact on disparity estimation and virtual perspective in the late processing of video compression. Therefore, color correction technology is a key technique of the multi-view video and has a wide application value. This paper focuses on the research of color correction technology in multi-view video, and the main achievement focus on the followings:1. A multi-view video color correction method based on 3D Gaussian Mixture Models is proposed. For the clustering analysis of source image and target image, the three-dimensional Gaussian Mixture Models is performed, while the center vector is obtained. Then, the algorithm calculates the minimum Euclidean distance between each pixel and the center vector to segment the image, histogram matching correction is followed. Experimental results show that the proposed method avoids the excessive correction and gets a good effect comparing with the traditional histogram matching algorithm.2. A fast parameter estimation method is designed to estimate the parameter values in the multi-view video color correction method space.Experimental results show that the method can get an ideal correction effect.3. The popular video quality assessment methods are analyzed and compared. Based on that, a evaluation method using local regional is studied. Results show that the proposed color correction method and obtain a good effect.