View Sequence Coding using Warping-based Image Alignment for Multi-view Video

Multi-view video sequences are recorded by multiple cameras simultaneously. In view direction the images captured by different cameras in the same time instant form the view image sequences. The view sequence represents the geometry changes of the scene in space. The compression of view sequence provides the great contribution to the whole multi-view video compression. This paper discusses the view sequence coding (VSC) and proposes a warping-based alignment technology to improve the compression performance. Proposed method caters to the traits of view sequence and adopts the affine model to warp reference image. Image inpainting making use of information from adjacent warped image is proposed to process the missing regions formed while warping. The aligned multi-hypothesis disparity prediction is employed to further improve the prediction effects. The experiment shows that the coding efficiency of the proposed scheme keeps head and shoulders above all the intra coding in simulcast coding scheme and the proposed scheme also improves the coding efficiency by up 0.3 ~ 0.5 dB at low bit rates when compared with the standard coding method.