Comparison of Fourier transform, windowed Fourier transform,
and wavel et transform methods for phase calculation
at discontin uities in fringe projection profilometry
Zonghua Zhang
a,
n
, Zhao Jing
a
, Zhaohui Wang
a
, Dengfeng Kuang
b
a
School of Mechanical Engineering, Hebei University of Technology, Tianjin 300130, China
b
Institute of Modern Optics, the Key Laboratory of Optical Information Science and Technology of the Education Ministry of China, Nankai University, Tianjin 300071, China
article info
Available online 26 March 2012
Keywords:
Fringe projection profilometry
Discontinuities measurement
Fourier transform
Windowed Fourier transform
Wavelet transform
abstract
Phase demodulation techniques from one fringe pattern have been widely studied because it can
measure dynamic objects by capturing single image. These techniques mainly include Fourier trans-
form (FT), windowed Fourier transform (WFT), and wavelet transform (WT). FT has been widely used to
demodulate phase information from single deformed fringe pattern on smooth objects. However, for
objects having discontinuities and/or large slopes, FT cannot obtain correct phase at the edges because
of its global processing. WFT and WT have been applied to nonstationary fringe pattern analysis. Since
local fringe information used to extract phase information, WFT and WT are better than FT for phase
calculation at discontinuities and/or slopes. In this paper, we discuss the pro and con of the three
methods on phase calculation at discontinuities and/or slopes. Simulated and experimental data are
tested at edges in order to confirm which method is appropriate to measure objects having
discontinuities by using one-frame fringe pattern acqu isition method.
& 2012 Elsevier Ltd. All rights reserved.
1. Introduction
Non-contact optical full-field 3D sensing techniques are
active research fields in machine vision, industrial inspection,
reverse engineering, physical profiling, biomedical testing, and
many other fields [1,2]. Among them, phase calculation-based
fringe projection techniques are widely studied because of
the advantages of fast acquisition and accurate data retrieval.
Two traditional methods for phase calculation are phase-shift-
ing profilometry (PSP) [3] and Fourier transform profilometry
(FTP) [4]. PSP method processes the fringe patterns pixel by pixel,
but it needs at least three fringe patterns having constant phase
shift in between and this technique is sensitive to noise and
environment. Therefore, it is hard to measure moving objects by
using PSP-based methods.
Fourier transform (FT) method has been widely used to
demodulate phase information from single deformed fringe
pattern on smooth objects, so it has been applied to such fields
as fast acquisition 3D shape of unsteady objects. However, for
objects having discontinuities and/or large slopes, FT cannot
obtain correct phase at the edges because of its global processing.
All frequency components in frequency domain are decided by
the entire spatial information, so the higher frequency compo-
nents at the edges may distribute in the whole wrapped phase
map. The obtained phase and then shape data are incorrect
because of the frequency leakage.
In order to demodulate nonstationary and noisy fringe pat-
terns, recently, windowed Fourier transform (WFT) [5] and
wavelet transform (WT) [6] profilometry are proposed. Since they
use local fringe information to extract phase information, the two
techniques are better than FT for phase calculation at disconti-
nuities and/or slopes. Most recently, Huang et al. compared FT,
WFT and WT methods for phase extraction from a single fringe
pattern on smooth surface [7]. Although they pointed out that all
the three transform-based algorithms perform poorly at edges,
there are no further researches investigated regarding the differ-
ences among FT, WFT and WT algorithms. Gorthi et al. mainly
discussed the effects of non-periodicity of fringe pattern on errors
of phase calculation and some qualitative comparisons between
FT and WFT at discontinuities [8]. Li et al. investigated different
wavelets (1D and 2D) for phase recovery on a hemispherical
object [9]. Quan et al. mainly gave theoretical description,
simulated and experimental results of phase-shifting, FT, WFT
and WT algorithms, and then briefly discussed their performance
on phase calculation at discontinuities [10]. However, no one has
given the quantitative evaluation and analysis at discontinuities
of a single fringe pattern among FT, WFT and WT algorithms.
Contents lists available at SciVerse ScienceDirect
journal homepage: www.elsevier.com/locate/optlaseng
Optics and Lasers in Engineering
0143-8166/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.optlaseng.2012.03.004
n
Corresponding author. Tel./fax: þ 86 22 60204189.
E-mail address: zhzhangtju@hotmail.com (Z. Zhang).
Optics and Lasers in Engineering 50 (2012) 1152–1160