Page 148 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Chapter 5

Warping-Based Motion

Estimation Techniques

5.1 Overview

As already discussed, one way to achieve higher coding e ciency is to
improve the performance of the motion estimation and compensation processes.
This can be done by using advanced motion estimation and compensation tech-
niques. This chapter concentrates on an advanced technique called warping-
based motion estimation. Since the early 1990s, this technique has attracted
attention in the video coding community as an alternative to (or rather as a
generalization of) conventional block-matching methods.
Section 5.2 reviews warping-based motion estimation techniques. Various
aspects of such techniques, like the shape of patches, the type of meshes,
the spatial transformation, the continuity of the motion (eld, the direction of
node tracking, the node-tracking algorithm, the motion compensation method,
and the transmitted motion overhead, are considered and compared. Section 5.3
compares the performance of warping-based methods to that of block-matching
methods. In particular, the section investigates the e ciency of warping-based
methods at very low bit rates. The chapter concludes with a discussion in
Section 5.4.

5.2 Warping-Based Methods: AReview

Motion estimation (ME) can be de(ned as a process that divides the current
frame, f c , into regions and that estimates for each region a set of motion
parameters, {a i }, according to a motion model. The motion compensation
(MC) process then uses the estimated motion parameters and the motion model
ˆ
to synthesize a prediction, f, of the current frame from a reference frame, f r .
c
125