Multiple description video coding using coefficients ordering and interpolation

• Authors: Nicola Conci;Francesco G. B. De Natale
• Venue: Image Communication
• Year: 2007

A Multiple Description Video Codec is presented, suitable for high-quality digital video transmission. The method works in the DCT domain, and associates to each descriptor two data sources: a subset of the original transform coefficients, and side information concerning the position and magnitude of the remaining coefficients. When one or more descriptors are (partially) lost, the side information is used at the decoder to interpolate the missing coefficients, thus achieving a better reconstruction. The effectiveness of the proposed approach is greatly enhanced by the use of a JPEG-like syntax and an efficient 3D-VLC in the encoding of the overhead information. This choice makes also easier and natural the integration of the MDC scheme in most DCT-based standard encoders such as M-JPEG and MPEG-1/2. With respect to previous approaches to MDC, the proposed strategy is characterized by a limited overhead and a reduced impact in terms of computational complexity. Furthermore, the coefficient selection strategy allows splitting the video stream into an arbitrary number of descriptors, overcoming a major limitation of most MDC schemes. These features make the proposed method suitable for real-time applications of video delivery, especially in wireless environments where the computational complexity is usually performed by low-power devices. As far as the reconstruction quality is concerned, extensive experimental sessions demonstrated that the loss of one descriptor produces a slight degradation in a two-descriptors scheme, while it is almost seamless in a three-descriptors scheme.