Presenter

Mr Jan Hanca - ETRO, Vrije Universiteit Brussel [Email]

Abstract

Ranging from surveillance and environmental tracking to healthcare monitoring and industrial supervision, Wireless Sensor Networks (WSN) have proven to be a useful solution to many application domains. Additionally, the increasing availability of lowcost CMOS cameras has created the opportunity to extend the sensing capabilities of sensor nodes, yielding more accurate description of the area under observation. In this regard, a video transmission scheme that allows error-resilient communication at very low encoding complexity is a key component of each visual WSN. While the performance of traditional predictive video coding schemes is reflected in the high operational cost of the encoder, Distributed Video Coding (DVC) has shown its potential to achieve state-of-the-art compression efficiency while maintaining low computational complexity of the compression system. Hence, this makes DVC systems appealing for battery-powered sensing devices, which is of critical importance in applications such as healthcare monitoring or capsule endoscopy. However, many DVC architecture designs overlook hardware constraints, and this renders them unsuitable for practical implementations. Selecting a proper video coding architecture for WSN applications requires thus the right balance between the advantages and disadvantages of both coding paradigms.
This thesis proposes an in-depth analysis of two video coding paradigms applied on low- and extremely low-resolution video data. First, the number of simplifications applied to conventional predictive compression mechanisms is introduced. An exhaustive evaluation and comparison against standardized methods reveals the impact on the compression performance of reducing the computational complexity of each coding tool. Furthermore, error resilient coding mechanisms offering protection against transmission errors on the wireless communication channel are devised.
Next, a novel DVC architecture that offers highly effective wireless communication in real-world visual WSNs is proposed. The design takes into account the severe computational and memory constraints imposed by practical hardware implementations. Most importantly, it suppresses the presence of a feedback communication channel, which is typical for DVC architectures. The capability of joint coding and robust transmission in error-prone environments incorporated into this design is compared against independent coding and protection mechanisms introduced for conventional predictive coding paradigms. Finally, the encoding performance, computational complexity and power consumption of both systems implemented in the existing hardware is set side-by-side and thoroughly compared, revealing advantages and trade-offs for each coding paradigm

Short CV

MSc Politechnika Poznanska, Poland, 2010