Resource aware real-time stream adaptation of MPEG-4 video in constrained bandwidth networks

In this paper, we propose a novel method for real- time stream adaptation of live MPEG-4 Visual advanced simple profile (ASP) and simple profile (SP) bit streams to limited and varying network bandwidth. The proposed method operates by dropping higher order AC DCT coefficients present in blocks. Unlike frame dropping, the proposed method delivers finer granularity during adaptation and therefore smaller resource under-utilization. Furthermore, to optimize the resource utilization and consequently the perceptive video quality, the method attempts to represent the adapted block with the least possible number of bits. The low computation requirements of the proposed method qualifies its application to adaptation of live streams. Additionally, the method can be applied for other constrained resources as well (e.g. CPU bandwidth). Implementation and case study of the proposed method show that, on an average, a bit rate reduction of up to 50% and a decoding time speedup up to 20% are achieved when compared with that of the original bitstream. Results also affirm the efficiency in resource usage.