Improved rate control via dynamic frame-rate controlling and Kalman filtering for low-delay video coding

Abstract

This study proposes a rapid rate-control framework for low-delay video coding. The proposed framework includes a fast frame-rate controller, a frame-level bit-allocation operator, and an adaptive rate-quantization (R-Q) model. The frame-rate controller and the frame-level bit-allocation operator determine the encoding frame rate per time slot and the bit target per frame, which specify motion continuity and consistency of quality, respectively. When a cluster of macroblocks (MBs) in an encoding frame is set by a progressive union process, the adaptive R-Q model is modified by the Kalman filter. The newest R-Q characteristic linear curves of the adaptive R-Q model can be used to obtain the MB-level quantization parameters in a new cluster of MBs (COB). The proposed new framework exhibits higher objective and perceptual qualities than the TMN8 rate control, using a TMN11 frame-rate control module, with a lower computational load.