Improving the Prediction Efficiency for Multi-View Video Coding Using Histogram Matching

Abstract. Applications for video data recorded with a setup of several cameras are currently attracting increasing interest. For such multi-view sequences, efficient coding is crucial to handle the enormous amount of data. However, significant luminance and chrominance variations between the different views, which often originate from imperfect camera calibration, are able to reduce the coding efficiency and the rendering quality. In this paper, we suggest the usage of histogram matching to compensate these differences in a pre-filtering step. After a description of the proposed algorithm, it is explained how histogram matching can be applied to multi-view video data. The effect of histogram matching on the coding performance is evaluated by statistically analysing prediction from temporal as well as from spatial references. For several test sequences, results are shown which indicate that the amount of spatial prediction across different camera views can be increased by applying histogram matching. Index Terms – multi-view video, video coding, image filtering