Multiple Description Image Coding With Hybrid Redundancy

This paper introduces a novel and effective multiple description image coding method by combination of polyphase downsampling and multiple description scalar quantizers. Different from previous works, the natural redundancy in images is fully exploited by polyphase downsampling on the spatial domain of images to help improving reconstruction quality by effective error concealment. This approach is also flexible for generating more than two descriptions with very low complexity and for distributing redundancy among them to combat possible erasures. Simulation results on real images demonstrate high robustness of the proposed scheme for highly lossy networks with packet loss up to 87.5%