Compression of side information

We consider the problem of data compression with side information at the decoder and propose novel algorithms for quantization of the side information. These methods build quantizers that minimize the rate at which the source can be encoded with a constraint on the entropy of the quantized side information. We study relationships between our methods and those of context quantization used for image and video coding, and agglomerative information bottleneck in document classification. Our algorithms are shown to perform well on simple experiments and are applicable to a wide range of distributed multimedia compression problems.