Low delay video coding

Analogue wireless cameras have been employed for decades, however they have not become an universal solution due to their difficulties of set up and use. The main problem is the link robustness which mainly depends on the requirement of a line-of-sight view between transmitter and receiver, a working condition not always possible. Despite the use of tracking antenna system such as the Portable Intelligent Tracking Antenna (PITA [1]), if strong multipath fading occurs (e.g. obstacles between transmitter and receiver) the picture rapidly falls apart. Digital wireless cameras based on Orthogonal Frequency Division Multiplexing (OFDM) modulation schemes give a valid solution for the above problem. OFDM offers strong multipath protection due to the insertion of the guard interval; in particular, the OFDM-based DVB-T standard has proven to offer excellent performance for the broadcasting of multimedia streams with bit rates over 10 Mbps in difficult terrestrial propagation channels, for fixed and portable applications. However, in typical conditions, the latency needed to compress/decompress a digital video signal at Standard Definition (SD) resolution is of the order of 15 frames, which corresponds to ≃ 0.5 sec. This delay introduces a serious problem when wireless and wired cameras have to be interfaced. Cabled cameras do not use compression, because the cable which directly links transmitter and receiver does not impose restrictive bandwidth constraints. Therefore, the only latency that affects a cable cameras link system is the on cable propagation delay, almost not significant, when switching between wired and wireless cameras, the residual latency makes it impossible to achieve the audio-video synchronization, with consequent disagreeable effects. A way to solve this problem is to provide a low delay digital processing scheme based on a video coding algorithm which avoids massive intermediate data storage. The analysis of the last MPEG based coding standards puts in evidence a series of problems which limits the real performance of a low delay MPEG coding system. The first effort of this work is to study the MPEG standard to understand its limit from both the coding delay and implementation complexity points of views. This thesis also investigates an alternative solution based on HERMES codec, a proprietary algorithm which is described implemented and evaluated. HERMES achieves better results than MPEG in terms of latency and implementation complexity, at the price of higher compression ratios, which means high output bit rates. The use of HERMES codec together with an enhanced OFDM system [2] leads to a competitive solution for wireless digital professional video applications.

Related material


EPFL authors