Multi-View Codec using Disparity Compensation

Multi-view video applications are getting popular these days with newly introduced 3D displays. Compression of these sequences are mostly done by applying H.264 codec to each video sequence separately. Since there is redundancy between different camera views, coding gain can be improved by exploiting those redundancies with the use of disparity estimation/compensation. In this paper, we are introducing H.264 based multi-view video codec with additional disparity estimation/compensation. For coding and speed optimization, we have modified the buffering structure and introduced different referencing modes. Our codec outperforms independent H.264 coding when cameras are closely located. If cameras are sparsely located, coding gain can still be improved depending on the video characteristics