Image Deformation Using Moving Least Squares
Scott Schaefer
∗
Texas A&M University
Travis McPhail
†
Rice University
Joe Warren
‡
Rice University
(a) (b) (c) (d)
Figure 1: Deformation using Moving Least Squares. Original image with control points shown in blue (a). Moving Least Squares deforma-
tions using affine transformations (b), similarity transformations (c) and rigid transformations (d).
Abstract
We provide an image deformation method based on Moving Least
Squares using various classes of linear functions including affine,
similarity and rigid transformations. These deformations are real-
istic and give the user the impression of manipulating real-world
objects. We also allow the user to specify the deformations using
either sets of points or line segments, the later useful for control-
ling curves and profiles present in the image. For each of these
techniques, we provide simple closed-form solutions that yield fast
deformations, which can be performed in real-time.
CR Categories: I.3.5 [Computer Graphics]: Computational Ge-
ometry and Object Modeling—Boundary representations; Curve,
surface, solid, and object representations; Geometric algorithms,
languages, and systems
Keywords: Deformations, moving least squares, rigid transforma-
tions
1 Introduction
Image deformation has a number of uses from animation, to mor-
phing [Smythe 1990] and medical imaging [Warren et al. 2003].
To perform these deformations the user selects some set of han-
dles to control the deformation. These handles may take the form
of points [Bookstein 1989], lines [Beier and Neely 1992], or even
polygon grids [MacCracken and Joy 1996]. As the user modifies
∗
email: sschaefe@rice.edu
†
email:tjice@rice.edu
‡
email:jwarren@rice.edu
the position and orientation of these handles, the image should de-
form in an intuitive fashion.
We view this deformation as a function f that maps points in the
undeformed image to the deformed image. Applying the function
f to each point v in the undeformed image creates the deformed
image. Now consider an image with a set of handles p that the user
moves to new positions q. For f to be useful for deformations it
must satisfy the following properties:
• Interpolation: The handles p should map directly to q under
deformation. (i.e; f (p
i
) = q
i
).
• Smoothness: f should produce smooth deformations
• Identity: If the deformed handles q are the same as the p, then
f should be the identity function. (i.e; q
i
= p
i
⇒ f (v) = v).
These properties are very similar to those used in scattered data
interpolation. The first two properties simply state that the func-
tion f interpolates the scattered data values and is smooth. The last
property is sometimes referred to as linear precision in the approxi-
mation field. It states that if data is sampled from a linear function,
then the interpolant reproduces that linear function. Given these
similarities, it comes as no surprise that many deformation meth-
ods borrow techniques from scattered data interpolation.
Previous Work
Previous work on image deformation has focused on specify-
ing deformations using different types of handles. Grid-based
techniques such as free-form deformations [Sederberg and Parry
1986; Lee et al. 1995] parameterize the image using bivariate cubic
splines to create C
2
deformations. Typically these methods require
aligning grid lines corresponding to the control points of the spline
with features of the image, which can be cumbersome for the user.
Beier et al. [Beier and Neely 1992] improve upon these grid-
based techniques and allow the user to specify the deformation
using sets of lines. This method is based on Shepard’s inter-
polant [Shepard 1968] and creates smooth deformations. However,
the authors note that their method produces complicated warps that