Statistical Framework for Video Decoding Complexity Modeling and Prediction

摘要: Video decoding complexity modeling and prediction is an increasingly important issue for efficient resource utilization in a variety of applications, including task scheduling, receiver-driven shaping, adaptive dynamic voltage scaling. In this paper we present novel view problem based on statistical framework perspective. We explore the structure (clustering) execution time required by each video decoder module (entropy decoding, motion compensation, etc.) conjunction with features that are easily extractable at encoding (representing properties module's input source data). For purpose, employ Gaussian mixture models (GMMs) expectation-maximization algorithm to estimate joint execution-time-feature probability density function (PDF). A training set typical sequences used purpose offline estimation process. The obtained GMM representation new predict these sequences. Several approaches discussed compared. potential mismatch between content addressed online joint-PDF re-estimation. An experimental comparison performed evaluate different compare proposed scheme related schemes from literature. usefulness complexity-prediction demonstrated application rate-distortion-complexity optimized decoding.