Generalized multiple description coding with correlating transforms

Multiple description ND) coding is source coding in which several descripti ons of the source are produced such that various reconstruction qualities a re obtained from different subsets of the descriptions. Unlike multiresolut ion or layered source coding, there is no hierarchy of descriptions; thus, MD coding is suitable for packet erasure channels or networks without prior ity provisions. Generalizing work by Orchard, Wang, Vaishampayan, and Reibm an, a transform-based approach is developed for producing AT descriptions o f an N-tuple source, M less than or equal to N. The descriptions are sets o f transform coefficients, and th transform coefficients of different descri ptions are correlated so that missing coefficients can be estimated. Severa l transform optimization results are presented for memoryless Gaussian sour ces, including a complete solution of the N = 2, M = 2 case with arbitrary weighting of the descriptions. The technique is effective only when indepen dent components of the source have differing variances. Numerical studies s how that this method performs well at low redundancies, as compared to unif orm MD scalar quantization.