Abstract
The two major subpixel registration algorithms, currently being used in subset-based digital image correlation, are the classic Newton-Raphson (FA-NR) algorithm with forward additive mapping strategy and the recently introduced inverse compositional Gauss-Newton (IC-GN) algorithm. Although the equivalence of these two algorithms has been proved in existing studies, practical implementations of the two subpixel registration algorithms do involve differences, and therefore lead to different performance. In the present work, detailed theoretical error analyses of the two algorithms are performed. Based on the simple sum of squared difference criterion and the practical first-order shape function, analytic formulae that can quantify both the bias error (systematic error) and the variability (random error) in the displacements measured by IC-GN and FA-NR algorithms with various interpolation methods (i.e., cubic convolution interpolation, cubic polynomial interpolation, cubic B-spline interpolation and quintic B-spline interpolation) are derived. It is shown that, compared with FA-NR algorithm, IC-GN algorithm leads to reduced bias error in displacement estimation by eliminating noise-induced bias error, and gives rise on the average to smaller random errors in displacement estimation in the cases of high noise levels or using small subsets. Numerical tests with precisely controlled subpixel displacements confirm the correctness of the theoretical derivations. The results reveal that IC-GN algorithm outperforms the classic FA-NR algorithm not only in terms of computational efficiency, but also in respect of subpixel registration accuracy and noise-proof performance, and is strongly recommended as a standard subpixel registration algorithm for practical DIC applications instead of FA-NR algorithm.
Similar content being viewed by others
Notes
It should be noted that classic FA-NR algorithm normally use the intensity gradients of the updated deformed subsets to build up the Hessian matrix in each iteration. Although several works recommended [2, 36] to use the gradients of the reference image to enhance computational efficiency of FA-NR algorithm, which is claimed to be approximately equal to IC-GN algorithm in single iteration. Such a practice, namely without updating the Hessian matrix as required in FA-NR algorithm [37], generally leads to increased iteration number or even a failure in convergence.
It is important to emphasize here that the incremental warp with a deformation vector △p should be inverted and composite with the initial guess p = (u, u x, u y, v, v x, v y) using equation (3) to determined the updated deformation parameter vector p’(u’, u’x, u’y, v’, v’x, v’y). By definition, the displacement errors should be p e = p’- p . However, since the u xe, u ye, v xe, v ye estimated by equation (4) are small values approaching zero. Thus, displacement errors vector can be estimated as u e ≅ Δu, v e ≅ Δv.
References
Pan B, Qian K, Xie H, Asundi A (2009) Two-dimensional digital image correlation for in-plane displacement and strain measurement: a review. Meas Sci Tech 20(6):062001
Sutton MA, Orteu JJ, Schreier H (2009) Image correlation for shape, motion and deformation measurements: basic concepts, theory and applications. Springer Science & Business Media
Pan B (2011) Recent progress in digital image correlation. Exp Mech 51:1223–1235
Pan B, Xie HM, Xu BQ, Dai FL (2006) Performance of sub-pixel registration algorithms in digital image correlation. Meas Sci Tech 17:1615–1621
BruckH A, McNeil SR, Sutton MA, Peters WH (1989) Digital image correlation using Newton-Raphson method of partial differential correction. Exp Mech 29:261–267
Pan B, Li K (2011) A fast digital image correlation method for deformation measurement. Opt Lasers Eng 49(7):841–847
Baker S, Mattews I (2001) Equivalence and efficiency of image alignment algorithms. Proc IEEE Conf Comput Vis Pattern Recognit 56:1090–1097
Baker S, Mattews I (2004) Lucas-Kanade 20 years on: a unifying framework. Int J Comput Vision 56:221–255
Pan B, Li K, Tong W (2013) Fast, robust and accurate digital image correlation calculation without redundant computations. Exp Mech 53:1277–1289
Gao Y, Cheng T, Su Y, Xu X, Zhang Y, Zhang Q (2015) High-efficiency and high-accuracy digital image correlation for three-dimensional measurement. Opt Lasers Eng 65:73–80
Xu J, Moussawi A, Gras R, Lubineau G (2014) Using image gradients to improve robustness of digital image correlation to non-uniform illumination: effects of weighting and normalization choices. Exp Mech 55(5):963–979
Pan B, Tian L (2015) Superfast robust digital image correlation analysis using parallel computing. Opt Eng 54(3):034106
Shao X, Dai X, He X (2015) Noise robustness and parallel computation of the inverse compositional Gauss–Newton algorithm in digital image correlation. Opt Lasers Eng 71:9–19
Jiang ZY, Qian KM, Miao H, Yang JL, Tang LQ (2015) Path-independent digital image correlation with high accuracy, speed and robustness. Opt Lasers Eng 65:93–102
Pan B, Tian L, Song X (2016) Real-time, non-contact and targetless measurement of vertical deflection of bridges using off-axis digital image correlation. NDT E Int 79:73–80
Blaber J, Adair B, Antoniou A (2015) Ncorr: open-source 2D digital image correlation Matlab software. Exp Mech 55(6):1105–1122
Pan B, Wang B, Wu DF, Lubineau G (2014) An efficient and accurate 3D displacement tracking algorithm for digital volume correlation. Opt Lasers Eng 58:126–135
Wang ZY, Li HQ, Tong JW, Ruan JT (2007) Statistical analysis of the effect of intensity pattern noise on the displacement measurement precision of digital image correlation using self-correlated images. Exp Mech 47(5):701–707
Pan B, Xie HM, Wang ZY, Qian KM, Wang ZY (2008) Study of subset size selection in digital image correlation for speckle patterns. Opt Express 16(10):7037–7048
Wang YQ, Sutton MA, Bruck HA et al (2009) Quantitative error assessment in pattern matching: effects of intensity pattern noise, interpolation, strain and image contrast on motion measurements. Strain 45(2):160–178
Pan B (2013) Bias error reduction of digital image correlation using Gaussian pre-filtering. Opt Lasers Eng 51(10):1161–1167
Blaysat B, Grediac M, Sur F (2016) On the propagation of camera sensor noise to displacement maps obtained by DIC-an experimental study. Exp Mech 1–26. doi: 10.1007/s11340-016-0130-9
Schreier HW, Braasch JR, Sutton MA (2000) Systematic errors in digital image correlation caused by intensity interpolation. Opt Eng 9(11):2915–2921
Blaysat B, Grédiac M, Sur F (2016) Effect of interpolation in noise propagation from images to DIC displacement maps. Int J Numer Methods Eng. doi:10.1002/nme.5212
Su Y, Zhang Q, Gao Z, Xu X, Wu X (2015) Fourier-based interpolation bias prediction in digital image correlation. Opt Express 23(15):19242–19260
Baldi A, Bertolino F (2015) Experimental analysis of the errors due to polynomial interpolation in digital image correlation. Strain 51(3):248–263
Schreier HW, Sutton MA (2002) Systematic errors in digital image correlation due to undermatched subset shape functions. Exp Mech 42(3):303–310
Yu L, Pan B (2015) The errors in digital image correlation due to overmatched shape functions. Meas Sci Tech 26(4):045202
Wang B, Pan B (2015) Random errors in digital image correlation due to matched or overmatched shape functions. Exp Mech 55(9):1717–1727
Pan B, Lu ZX, Xie HM (2010) Mean intensity gradient: an effective global parameter for quality assessment of the speckle patterns used in digital image correlation. Opt Lasers Eng 48(4):469–477
Bornert M, Brémand F, Doumalin P et al (2009) Assessment of digital image correlation measurement errors: methodology and results. Exp Mech 49(3):353–370
Wang Y, Lava P, Reu P, Debruyne D (2015) Theoretical analysis on the measurement errors of local 2D DIC: part I temporal and spatial uncertainty quantification of displacement measurements. Strain. doi:10.1111/str.12173
Wang Y, Lava P, Reu P, Debruyne D (2016) Theoretical analysis on the measurement errors of local 2D DIC: part II assessment of strain errors of the local smoothing method–approaching an answer to the overlap question. Strain. doi:10.1111/str.12174
Tong W (2011) Subpixel image registration with reduced bias. Opt Lett 36(5):763–765
Pan B, Xie HM, Wang ZY (2010) Equivalence of digital image correlation criteria for pattern matching. Appl Opt 49:5501–5509
Besnard G, Hild F, Roux S (2006) “Finite-element” displacement fields analysis from digital images: application to Portevin-Le Châtelier bands. Exp Mech 46(6):789–803
Tong W (2013) Formulation of Lucas-Kanade digital image correlation algorithms for non-contact deformation measurements: a review. Strain 49(4):313–334
Pan B (2014) An evaluation of convergence criteria for digital image correlation using inverse compositional Gauss–Newton algorithm. Strain 50(1):48–56
Acknowledgments
This work is supported by the National Natural Science Foundation of China (Grant nos. 11272032, 11322220 and 11427802), Beijing Nova Program (xx2014B034), China Aerospace Science and Technology Innovation Fund Project, and the Fundamental Research Funds for the Central Universities
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A: Rigorous Mathematical Derivation of the Variance of Displacement Error for IC-GN Algorithm
By taking imperfect intensity interpolation into consideration, more general formula regarding the variance of deformation error vector can be deduced for IC-GN algorithm as follows
where the general rules of variance calculation, i.e., Var(XY) = [E(X)]2Var(Y) + [E(Y)]2Var(X) + Var(X)Var(Y), are utilized.
According to equation (B3) in Ref. [32], the variance of deformation error vector can be simplified as
It can be clearly seen that equation (A2) can be simplified to equation (13) by omitting the second term and removing interpolation error. Due to the relatively low SD of image noise, the following approximation can be obtained based on the use of the Barron operator
where (Δx i , Δy j ) is the fractional part of interpolation position in deformed subset; N is the subset size used in DIC. For the four aforementioned interpolation methods in 1D case, K(Δx i , Δy j ) is the sum-of-squares of the coefficients for the intensity, which can be expressed as
For clarity, function K(Δx i ) is plotted as a function of subpixel positions for the four different interpolation methods in Fig. 9(a).
Appendix B: Theoretical Error Assessment of DIC with FA-NR Algorithm
Similar to equation (6), the SSD criterion for DIC using FA-GN algorithm can be written as
Substituting the first-order Taylor expansion of g′(x ′ i , y ′ j ) into SSD function, one can obtain
Minimization of the C SSD function with respect to p e gives
where H can be written as
with \( \nabla {g}^{\prime}\left({x}_i^{\prime },{y}_j^{\prime}\right)\cdot \frac{\partial \mathbf{W}}{\partial \mathbf{p}}=\left({g}_x^{\prime },{g}_x^{\prime}\varDelta {x}_i,{g}_x^{\prime}\varDelta {y}_j,{g}_y^{\prime },{g}_y^{\prime}\varDelta {x}_i,{g}_y^{\prime}\varDelta {y}_j\right) \).
Analogously, the variance matrix of deformation vector errors p e can be simplified as
Based on the above equation, the variance of u e can be extracted as
To better approximate practical experiment, intensity interpolation should be considered with respect to equation (B6). According to Ref [32], more general governing formula can be expressed as
Due to the relatively low SD of image noise, the following approximation can be explained
where (Δx i , Δy j ) is the fractional part of interpolation position in deformed subset. For the four aforementioned interpolation methods in 1D case, D(Δx i , Δy j ) is the sum-of-squares of the coefficients for the intensity gradient. Figure 9(b) plots function D(Δx i ) against the fractional part of interpolation position for the four different interpolation methods. Due to the concave quadratic distribution of D ‐ 1(Δx i , Δy j ) and K(Δx i , Δy j ), the quadratic shape of SD error in FA-NR algorithm is theoretically predictable.
Also, the mathematical expectation of deformation vector errors p e can be described as
The interpolation-induced bias error Eint(p e ) can be represented as
Similarly, the interpolation-induced bias error of displacement component u e can be extracted as
And the noise-induced bias error Enoise(p e ) can be further written as
where (Δx i , Δy j ) is the fractional part of interpolation position in deformed subset; W(Δx i , Δy j ) denotes weight function at certain interpolation position. For example, in the case of one-dimension, weight functions of four aforementioned interpolation strategies have the following forms
Specifically, the noise-induced bias error of displacement component u e can be extracted as
It can be intuitively seen that mean bias errors due to FA-GN algorithm have direct relationship with image noise level, thus confirming its noise susceptibility. Figure 9(c) plots the weight functions as a function of the fractional part of interpolation position for various interpolation strategies.
Rights and permissions
About this article
Cite this article
Pan, B., Wang, B. Digital Image Correlation with Enhanced Accuracy and Efficiency: A Comparison of Two Subpixel Registration Algorithms. Exp Mech 56, 1395–1409 (2016). https://doi.org/10.1007/s11340-016-0180-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11340-016-0180-z