A multi-view coding scheme for interactive applications

View switching methods and decoding complexity are important concerns for interactive multi-view video applications. A multi-view video coding scheme is presented that improves view switching performance by eliminating relationships among predictive frames, while achieving high compression efficiency by reducing the temporal correlations within one view and by reducing the spatial correlations among views by using shared key frames and spatial-temporal predictive coding. The scheme provides switching points at each P-frame without auxiliary data and can be readily implemented with the H.264/AVC codec. Tests show that the scheme achieves good compression efficiency and flexible inactivity for highly interactive applications.