Page 114 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Part II

Coding E
ciency

The radio spectrum is a limited and scarce resource. This puts very stringent
limits on the bandwidth available for a mobile channel. Given the enormous
amount of data generated by video, the use of e cient coding techniques is
vital.
One of the most important factors that decide the coding e ciency of a
video codec is the motion estimation and compensation technique. This part
contains three chapters. Chapter 4 covers some basic motion estimation meth-
ods. It starts by introducing some of the fundamentals of motion estimation.
It then reviews some basic motion estimation methods, with particular em-
phasis on the widely used block-matching methods. The chapter then presents
the results of a comparative study between the di"erent methods. The chapter
also investigates the e ciency of block-matching motion estimation at very
low bit rates, typical of mobile video communication. The aim is to decide if
the added complexity of this process is justi%able, in terms of an improved
coding e ciency, at such bit rates.
Chapter 5 investigates the performance of the more advanced warping-based
motion estimation methods. The chapter starts by describing a general warping-
based motion estimation method. It then considers some important parameters,
like the shape of the patches, the spatial transformation used, and the node-
tracking algorithm. The chapter then assesses the suitability of warping-based
methods for mobile video communications. In particular, the chapter compares
the coding e ciency and the computational complexity of such methods to
those of block-matching methods.
Chapter 6 investigates the performance of another advanced motion estima-
tion method, called multiple-reference motion-compensated prediction (MR-
MCP). The chapter starts by brie-y reviewing multiple-reference motion esti-
mation methods. It then concentrates on the long-term memory motion-com-
pensated prediction (LTM-MCP) technique. The chapter investigates the
prediction gains and the coding e ciency of this technique at very low bit
rates. The primary aim is to decide if the added complexity, increased motion
overhead, and increased memory requirements of this technique are justi%able
at such bit rates. The chapter also investigates the properties of multiple-
reference block motion %elds and compares them to those of single-reference
%elds.