Integrated Multi-view Compensation for Real Sense Video Interfaces

Multi-view video is a new multimedia service which provides immersion and realism using the multiple view channels. However, as the number of cameras increases, the enormous data volume generated by them inevitably calls for better video compression and processing algorithms. However, the multiple images taken at any certain time instants always have inconsistency among the views in their intensity, color, sharpness, and so on. It not only degrades the compression efficiency greatly but also causes subjective quality loss such as dizziness and unnaturalness in random view access. To solve this problem, in this paper, we propose a new multi-view compensation method which uses the Kalman filter. It estimates the illumination discrepancy parameters autonomously both in encoder and decoder. Therefore, our method needs not transmit any additional data for illumination discrepancy information.