Abstract:
Rapid advances in Micro-Electro-Mechanical Systems (MEMS)
technology have made possible the existence of tiny sensor nodes
equipped with sensing, communication and processing capabil-
ities. When sprayed in an area, they are capable of forming a
multi-hop wireless network known as Wireless Sensor Network
(WSN). Wireless Video Sensor Network (WVSN) - a type of
WSNs - comprises of sensor nodes that can capture, process
and communicate video frames. However, sensor nodes have
limited hardware resources while video processing and commu-
nication is a hardware intensive task i.e. requires large memory
and bandwidth and high-end processors. Video encoding is a
popular method used to reduce the communication overhead in
low-energy WVSN. However, it is an inherently complex pro-
cess and therefore results in high processing and energy over-
head. Several video encoding schemes have been proposed in
the recent past which are rather simple to execute. However, an
independent, unbiased and thorough empirical analysis of these
video encoding techniques has never been done before and hence
is a subject of this thesis.
In this work, we consider two widely used video encoding
paradigms for wireless sensor networks { Predictive Video Cod-
ing (PVC) and Distributed Video Coding (DVC). These two
paradigms are empirically evaluated using Intel-mote2 which is a
state-of-the-art WVSN platform. The analysis has been carried
out to characterize the performance of both paradigms in terms
of energy e±ciency, compression e±ciency and video distortion.
We perform experiments in multi-hop scenarios using video se-
quences of di®erent characteristics. Each video sequence has
been evaluated using both the inter-frame encoding and intra-
frame encoding. The empirical analysis reveals that none of the
existing solutions can be readily adopted in all possible scenar-
ios. Rather, the codec selection should be based on a particular
WVSN application and available resources. We also formulate
detailed guidelines to facilitate the selection process.
