RATE-DISTORTION QUANTIZATION FOR CONTEXT-ADAPTIVE VARIABLE LENGTH CODING (CAVLC)

摘要

In general, this disclosure provides techniques for quantization of the coefficients of video blocks in a manner that can achieve a desirable balance of rate and distortion. The described techniques may analyze a plurality of quantization levels associated with each individual coefficient to select the quantization level for the individual coefficients that results in a lowest coding cost. Since CAVLC does not encode each coefficient independently, the techniques may compute the coding costs for each of the candidate quantization levels associated with the individual coefficients based on quantization levels selected for previously quantized coefficients and estimated (or predicted) quantization levels for subsequent coefficients of a coefficient vector. The quantization levels for each of the coefficients are selected based on computed coding costs to obtain a set of quantized coefficients that minimize a rate-distortion model.