Distributed Internet-adaptive image compression

Summary form only given. The contribution of this paper is an extension of the pruned tree-structured hierarchical vector quantization (PTSHVQ) algorithm, with a focus on designing and benchmarking a complete Internet-aware system for the real-time delivery of compressed images. We extend the PTSHVQ algorithm to enable the encoder to use a combination of “coarse” and “fine” rate control mechanisms. The coarse rate control is achieved by switching between successive HVQ lookup tables, thereby changing the bit rate on a logarithmic scale, e.g., to give 4, 2, 1, 0.5 bpp, etc. The fine rate control is achieved by sub-setting into the address space of the current HVQ lookup table, e.g., generating 4, 3.5, 3, 2.5 bpp. The combination of these two strategies of switching between HVQ stages and subsetting the available range of addresses at each stage gives us a much larger dynamic range as well as finer calibration of operational bit rates than any existing HVQ system. Organizing the codebooks into binary tree structures by using the greedy BFOS algorithm guarantees locally optimal rate-distortion behavior at each operational bit rate. The function of the distributed transcoder is to enable the regulation of operational bandwidth over a heterogeneous network, thereby guaranteeing fairness in broadcast type applications