Skip to main content
Log in

Linear Time Illumination Invariant Stereo Matching

  • Published:
International Journal of Computer Vision Aims and scope Submit manuscript

Abstract

This paper proposes a new similarity measure that is invariant to global and local affine illumination changes. Unlike existing methods, its computational complexity is very low. When used for stereo correspondence estimation, its computational complexity is linear in the number of image pixels and disparity searching range. It also outperforms the current state of the art similarity measures in terms of accuracy on the Middlebury benchmark (with radiometric differences).

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11

Similar content being viewed by others

Notes

  1. The trained patch size for Census transform is \(19 \times 19\).

References

  • Brown, M., Burschka, D., & Hager, G. D. (2003). Advances in computational stereo. IEEE Transactions Pattern Analysis and Machine Intelligence, 25(8), 993–1008.

    Article  Google Scholar 

  • Chia, A., Zhuo, S., Gupta, R., Tai, Y., Cho, S., Tan, P., et al. (2011). Semantic colorization with internet images. ACM Transcation on Graphics, 30(6), 156:1–156:8.

    Google Scholar 

  • Cox, I., Roy, S., & Hingorani, S. (1995). Dynamic histogram warping of image pairs for constant image brightness. In ICIP.

  • De-Maeztu, L., Mattoccia, S., Villanueva, A., & Cabeza, R. (2011). Linear stereo matching. In ICCV, Nov 2011 (pp. 1708–1715).

  • Ding, Y., Xiao, J., & Yu, J. (2011). Importance filtering for image retargeting. In CVPR.

  • Egnal, G. (2000). Mutual information as a stereo correspondence measure, Technical Report MS-CIS-00-20. Computer and Information Science, University of Pennsylvania.

  • Goesele, M., Snavely, N., Curless, B., Hoppe, H., & Seitz, S. (2007). Multi-view stereo for community photo collections. In ICCV.

  • He, K., Rhemann, C., Rother, C., Tang, X., & Sun, J. (2011) A global sampling method for alpha matting. In CVPR, June 2011 (pp. 2049–2056).

  • He, K., Sun, J., & Tang, X. (2011). Single image haze removal using dark channel prior. IEEE Transactions Pattern Analysis and Machine Intelligence, 33(12), 2341–2353.

    Article  Google Scholar 

  • He, K., Sun, J., & Tang, X. (2013). Guided image filtering. IEEE Transactions Pattern Analysis and Machine Intelligence, 35(6), 1397–1409.

    Article  Google Scholar 

  • Heo, Y., Lee, K., & Lee, S. (2009). Mutual information-based stereo matching combined with sift descriptor in log-chromaticity color space. In CVPR, 2009 (pp. 445–452).

  • Heo, Y., Lee, K., & Lee, S. (2011). Robust stereo matching using adaptive normalized cross-correlation. IEEE Transactions Pattern Analysis and Machine Intelligence, 33(4), 807–822.

    Article  MathSciNet  Google Scholar 

  • Heo, Y., Lee, K., & Lee, S. (2013). Joint depth map and color consistency estimation for stereo images with different illuminations and cameras. IEEE Transactions Pattern Analysis and Machine Intelligence, 35(5), 1094–1106.

  • Hestenes, M., & Stiefel, D. (1952). Methods of conjugate gradients for solving linear systems. IEEE Transactions Pattern Analysis and Machine Intelligence, 49, 409–436.

    MathSciNet  MATH  Google Scholar 

  • Hirschmuller, H. (2008). Stereo processing by semiglobal matching and mutual information. IEEE Transactions Pattern Analysis and Machine Intelligence, 30(2), 328–341.

    Article  Google Scholar 

  • Hirschmuller, H., & Scharstein, D. (2009). Evaluation of stereo matching costs on images with radiometric differences. IEEE Transactions Pattern Analysis and Machine Intelligence, 31(9), 1582–1599.

    Article  Google Scholar 

  • Hosni, A., Rhemann, C., Bleyer, M., Rother, C., & Gelautz, M. (2013). Fast cost-volume filtering for visual correspondence and beyond. IEEE Transactions Pattern Analysis and Machine Intelligence, 35(2), 504–511.

    Article  Google Scholar 

  • Kim, J., Kolmogorov, V., & Zabih, R. (2003). Visual correspondence using energy minimization and mutual information. In ICCV.

  • Klaus, A., Sormann, M., & Karner, K. (2006). Segment-based stereo matching using belief propagation and a self-adapting dissimilarity measure. In ICPR, 2006 (pp. 15–18).

  • Levin, A., Lischinski, D., & Weiss, Y. (2008). A closed-form solution to natural image matting. IEEE Transactions Pattern Analysis and Machine Intelligence, 30(2), 228–242.

    Article  Google Scholar 

  • Scharstein, D., & Szeliski, R. (2002a) Middlebury stereo datasets. http://vision.middlebury.edu/stereo/data/.

  • Scharstein, D., & Szeliski, R. (2002b). A taxonomy and evaluation of dense two-frame stereo correspondence algorithms. International Journal of Computer Vision, 47, 7–42.

  • Snavely, N., Seitz, S. M., & Szeliski, R. (2006). Photo tourism: Exploring photo collections in 3d. IEEE Transactions Pattern Analysis and Machine Intelligence, 25(3), 835–846.

    Google Scholar 

  • Wang, L., & Yang, R. (2011). Global stereo matching leveraged by sparse ground control points. In CVPR, June 2011 (pp. 3033–3040).

  • Wang, L., Yang, R., & Davis, J. (2007). Brdf invariant stereo using light transport constancy. IEEE Transactions Pattern Analysis and Machine Intelligence, 29(9), 1616–1626.

    Article  Google Scholar 

  • Yang, Q., Wang, L., & Ahuja, N. (2010) A constant-space belief propagation algorithm for stereo matching. In CVPR, 2010 (pp. 1458–1465).

  • Yang, Q., Wang, L., Yang, R., Stewenius, H., & Nister, D. (2009). Stereo matching with color-weighted correlation, hierachical belief propagation and occlusion handling. IEEE Transactions Pattern Analysis and Machine Intelligence, 31(3), 492–504.

    Article  Google Scholar 

  • Yang, Q. (2015). Stereo matching using tree filtering. IEEE Transactions Pattern Analysis and Machine Intelligence, 37(4), 834–846.

    Article  Google Scholar 

  • Yoon, K.-J., & Kweon, I.-S. (2006). Adaptive support-weight approach for correspondence search. IEEE Transactions Pattern Analysis and Machine Intelligence, 28(4), 650–656.

    Article  Google Scholar 

  • Zabih, R., & Woodfill, J. (1994). Non-parametric local transforms for computing visual correspondence. In ECCV, 1994 (pp. 151–158).

  • Zhu, S., Zhang, L., & Jin, H. (2012). A locally linear regression model for boundary preserving regularization in stereo matching. In Proceedings of the 12th European conference on computer vision—Volume Part V, ser. ECCV’12 (pp. 101–115). Berlin: Springer-Verlag.

  • Zomet, A., & Peleg, S. (2002). Multi-sensor super resolution. In WACV.

Download references

Acknowledgments

This work was supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. CityU 21201914).

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Qingxiong Yang.

Additional information

Communicated by Masatoshi Okutomi.

Appendix

Appendix

1.1 Appendix 1: Derivation of Eq. 12

The matching cost measured from two corresponding pixels p and \(p'\) in two grayscale images \(I_L\) and \(I_R\) is:

$$\begin{aligned}&\sum _{q\in I_L}\mathcal {W}(p,q)\cdot (I_L(q)- a_p\cdot I_{R,\varDelta }(q)-b_p)^2\nonumber \\&\quad =\sum _{q\in I_L}\mathcal {W}(p,q)(a_p^2I_{R,\varDelta }(q)^2+2a_pb_pI_{R,\varDelta } (q)\nonumber \\&\qquad -2a_pI_L(q)I_{R,\varDelta }(q) -2b_pI_L(q)+b_p^2+I_L(q)^2)\nonumber \\&\quad =a_p^2\sum _{q\in I_L}\mathcal {W}(p,q)I_{R,\varDelta }(q)^2\!+\!2a_pb_p\sum _{q\in I_L}\mathcal {W}(p,q)I_{R,\varDelta }(q)\nonumber \\&\qquad -\,2a_p\sum _{q\in I_L}\mathcal {W}(p,q) I_L(q)I_{R,\varDelta }(q)\nonumber \\&\qquad -\,2b_{p}\sum _{q\in I_L}\mathcal {W}(p,q)I_L(q)+b_p^2+\sum _{q\in I_L}\mathcal {W}(p,q)I_L(q)^2\nonumber \\&\quad =a_p^2\cdot \mathcal {F}_{I_{R,\varDelta }\cdot I_{R,\varDelta }}(p)+2a_pb_p\cdot \mathcal {F}_{I_{R,\varDelta }}(p)\nonumber \\&\qquad -2a_p\cdot \mathcal {F}_{I_L\cdot I_{R,\varDelta }}(p)-\,2b_p\cdot \mathcal {F}_{I_{L}}(p)+b_p^2+\mathcal {F}_{I_L\cdot I_L}(p).\nonumber \\ \end{aligned}$$
(27)

1.2 Appendix 2: Derivation of Eq. 22

Similar to Eq. 4, we can extend Eq. 20 for color images as follows:

$$\begin{aligned} \tilde{A}\cdot \tilde{\mathcal {X}} =\tilde{B}. \end{aligned}$$
(28)

\(\tilde{\mathcal {X}}\) is defined in Eq. 21, and

$$\begin{aligned} \tilde{A}=\begin{pmatrix} \cdots &{}\mathcal {W}(p,q)\cdot I_{R,\varDelta }^1(q)&{}\cdots \\ \cdots &{}\mathcal {W}(p,q)\cdot I_{R,\varDelta }^2(q)&{}\cdots \\ \cdots &{}\mathcal {W}(p,q)\cdot I_{R,\varDelta }^3(q)&{}\cdots \\ \cdots &{}\mathcal {W}(p,q)\cdot 1&{}\cdots \\ \end{pmatrix}^T, \end{aligned}$$
(29)

and

$$\begin{aligned} \tilde{B}=\begin{pmatrix} \cdots &{}\mathcal {W}(p,q)\cdot I_L^1(q)&{}\cdots \\ \cdots &{}\mathcal {W}(p,q)\cdot I_L^2(q)&{}\cdots \\ \cdots &{}\mathcal {W}(p,q)\cdot I_L^3(q)&{}\cdots \\ \end{pmatrix}^T. \end{aligned}$$
(30)

The linear system presented in Eq. 28 can be rewritten as:

$$\begin{aligned} \tilde{A}^T\tilde{A}\cdot \tilde{\mathcal {X}} =\tilde{A}^T\tilde{B}, \end{aligned}$$
(31)

where

$$\begin{aligned}&\tilde{A}^T\tilde{A}\nonumber \\&\quad =\begin{pmatrix} \mathcal {F}_{I_{R,\varDelta }^1\cdot I_{R,\varDelta }^1}(p)&{}\mathcal {F}_{I_{R,\varDelta }^1\cdot I_{R,\varDelta }^2}(p)&{}\mathcal {F}_{I_{R,\varDelta }^1\cdot I_{R,\varDelta }^3}(p)&{}\mathcal {F}_{I_{R,\varDelta }^1}(p)\\ \mathcal {F}_{I_{R,\varDelta }^2\cdot I_{R,\varDelta }^1}(p)&{}\mathcal {F}_{I_{R,\varDelta }^2\cdot I_{R,\varDelta }^2}(p)&{}\mathcal {F}_{I_{R,\varDelta }^2\cdot I_{R,\varDelta }^3}(p)&{}\mathcal {F}_{I_{R,\varDelta }^2}(p)\\ \mathcal {F}_{I_{R,\varDelta }^3\cdot I_{R,\varDelta }^1}(p)&{}\mathcal {F}_{I_{R,\varDelta }^3\cdot I_{R,\varDelta }^2}(p)&{}\mathcal {F}_{I_{R,\varDelta }^3\cdot I_{R,\varDelta }^3}(p)&{}\mathcal {F}_{I_{R,\varDelta }^3}(p)\\ \mathcal {F}_{I_{R,\varDelta }^1}(p)&{}\mathcal {F}_{I_{R,\varDelta }^2}(p)&{}\mathcal {F}_{I_{R,\varDelta }^3}(p)&{}1\\ \end{pmatrix}\nonumber \\ \end{aligned}$$
(32)

and

$$\begin{aligned} \tilde{A}^T\tilde{B}= \begin{pmatrix} \mathcal {F}_{I_{R,\varDelta }^1\cdot I_L^1}(p)&{}\mathcal {F}_{I_{R,\varDelta }^1\cdot I_L^2}(p)&{}\mathcal {F}_{I_{R,\varDelta }^1\cdot I_L^3}(p)\\ \mathcal {F}_{I_{R,\varDelta }^2\cdot I_L^1}(p)&{}\mathcal {F}_{I_{R,\varDelta }^2\cdot I_L^2}(p)&{}\mathcal {F}_{I_{R,\varDelta }^2\cdot I_L^3}(p)\\ \mathcal {F}_{I_{R,\varDelta }^3\cdot I_L^1}(p)&{}\mathcal {F}_{I_{R,\varDelta }^3\cdot I_L^2}(p)&{}\mathcal {F}_{I_{R,\varDelta }^3\cdot I_L^3}(p)\\ \mathcal {F}_{I_L^1}(p)&{}\mathcal {F}_{I_L^2}(p)&{}\mathcal {F}_{I_L^3}(p)\\ \end{pmatrix}. \end{aligned}$$
(33)

1.3 Appendix 3: Derivation of Eq. 23

The matching cost for color images is:

$$\begin{aligned}&\sum _{{c1}=1}^3\sum _{q\in I_L}\mathcal {W}(p,q)\cdot (I_L^{c1}(q)-\sum _{{c2}=1}^3a_p^{{c1}{c2}}I_{R,\varDelta }^{c2}(q)-b_p^{c1})^2\nonumber \\&\quad =\sum _{{c1}=1}^3\sum _{q\in I_L}\mathcal {W}(p,q)\cdot ((I_L^{c1}(q))^2+(\sum _{{c2}=1}^3a_p^{{c1}{c2}}I_{R,\varDelta }^{c2}(q))^2\nonumber \\&\qquad +\,(b_p^{c1})^2-\,2b_p^{c1}I_L^{c1}(q)-2\sum _{{c2}=1}^3a_p^{{c1}{c2}}I_{R,\varDelta }^{c2}(q)I_L^{c1}(q)\nonumber \\&\qquad +\,2\sum _{{c2}=1}^3a_p^{{c1}{c2}}b_p^{c1}I_{R,\varDelta }^{c2}(q))\nonumber \\&\quad =\sum _{{c1}=1}^3(\sum _{q\in I_L}\mathcal {W}(p,q)\cdot I_L^{c1}(p)I_L^{c1}(p))\nonumber \\&\qquad +\sum _{{c3}=1}^3\sum _{{c2}=1}^3(\sum _{{c1}=1}^3 a_p^{c1c2}\cdot a_p^{c1c3})(\sum _{q\in I_L}\mathcal {W}(p,q)\nonumber \\&\qquad \cdot I_{R,\varDelta }^{c2}(p)I_{R,\varDelta }^{c3}(p))\nonumber \\&\qquad +\sum _{{c1}=1}^3(b_p^{c1})^2-2\sum _{{c1}=1}^3 b_p^{c1}(\sum _{q\in I_L}\mathcal {W}(p,q)\cdot I_L^{c1}(p))\nonumber \end{aligned}$$
$$\begin{aligned}&\qquad -2\sum _{{c1}=1}^3\sum _{{c2}=1}^3 a_p^{c1c2}(\sum _{q\in I_L}\mathcal {W}(p,q)\cdot I_L^{c1}(p)I_{R,\varDelta }^{c2}(p))\nonumber \\&\qquad +2\sum _{{c2}=1}^3\left( (\sum _{{c1}=1}^3a_p^{{c1}{c2}}\cdot b_p^{{c1}})\cdot (\sum _{q\in I_L}\mathcal {W}(p,q)\cdot I_{R,\varDelta }^{c2}(p))\right) \nonumber \\&\quad =\sum _{{c1}=1}^3\mathcal {F}_{I_L^{c1}\cdot I_L^{c1}}(p)\!+\!\sum _{{c3}=1}^3\sum _{{c2}=1}^3(\sum _{{c1}=1}^3 a_p^{c1c2}\cdot a_p^{c1c3}) \mathcal {F}_{I_{R,\varDelta }^{c2}\cdot I_{R,\varDelta }^{c3}}(p)\nonumber \\&\qquad +\,\sum _{{c1}=1}^3(b_p^{c1})^2 -2\sum _{{c1}=1}^3 b_p^{c1}\mathcal {F}_{I_L^{c1}}(p)\nonumber \\&\qquad -\,2\sum _{{c1}=1}^3\sum _{{c2}=1}^3 a_p^{c1c2}\mathcal {F}_{I_L^{c1}\cdot I_{R,\varDelta }^{c2}}(p)\nonumber \\&\qquad +\,2\sum _{{c2}=1}^3\left( (\sum _{{c1}=1}^3a_p^{{c1}{c2}}\cdot b_p^{{c1}})\cdot \mathcal {F}_{I_{R,\varDelta }^{c2}}(p)\right) \end{aligned}$$
(34)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Xu, J., Yang, Q., Tang, J. et al. Linear Time Illumination Invariant Stereo Matching. Int J Comput Vis 119, 179–193 (2016). https://doi.org/10.1007/s11263-016-0886-5

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11263-016-0886-5

Keywords

Navigation