Effective energy-based multi-view piecewise planar stereo

被引：0

作者：

Wang Y. ^{[1
]}

Wang W. ^{[1
]}

Zhu H. ^{[1
]}

Dong S. ^{[2
,3
]}

机构：

[1] School of Network Engineering, Zhoukou Normal University, Zhoukou

[2] School of Computer Science and Technology, Zhoukou Normal University, Zhoukou

[3] Department of Computer Science and Engineering, Washington University in St. Louis, Saint Louis, 63130, MO

来源：

Pattern Recognition and Image Analysis | 2016年 / 26卷 / 04期

关键词：

depth map; energy optimization; multi-view stereo; piecewise planar stereo; plane fitting;

D O I：

10.1134/S1054661816040209

中图分类号：

学科分类号：

摘要：

For piecewise planar scene modeling, many challenging issues still persist, in particular, how to generate sufficient candidate planes and how to assign an optimal plane for each spatial patch. To address these issues, we present a novel multi-view piecewise planar stereo method for the complete reconstruction. In our method, reconstruction is formulated as an energy-based plane labeling problem, where photo-consistency and geometric constraints are incorporated to a unified superpixel-level MRF (Markov Random Field) framework. To enhance the efficacy of the plane inference and optimization, an effective multi-direction plane sweeping with much restricted search space is carried out to generate sufficient and reliable candidate planes. Experiments show that our method can effectively handle many challenging factors (e.g., slant surfaces, textureless regions) and achieve satisfactory results. © 2016, Pleiades Publishing, Ltd.

引用

页码：726 / 733

页数：7

共 19 条

[1]

Peng Y., Li G., Wang R., Wang W., Stereo matching with space-constrained cost aggregation and segmentation- based disparity refinement, Proc. Three-Dimensional Image Processing, Measurement, (2015)

[2]

Veldandi M., Ukil S., Govindarao K., Robust segment-based stereo using cost aggregation, Proc. British Machine Vision Conf., pp. 1-11, (2014)

[3]

Gorelick L., Boykov Y., Veksler O., Ayed I.B., Delong A., Submodularization for binary pairwise energies, Proc. Computer Vision and Pattern Recogition, (2014)

[4]

Cigla C., Zabulis X., Alatan A.A., Region-based dense depth extraction from multi-view video, Proc. Int. Conf. on Image Processing (ICIP), pp. 213-216, (2007)

[5]

Furukawa Y., Curless B., Seitz S., Szeliski R., Manhattan-world stereo, Proc.Computer Vision and Pattern Recogition, pp. 1422-1429, (2009)

[6]

Furukawa Y., Ponce J., Accurate, dense, and robust multi-view stereopsis, IEEE Trans. Pattern Anal. Mach. Intellig., 32, pp. 1362-1376, (2009)

[7]

Micusik B., Kosecka J., Piecewise planar city 3D modeling from street view panoramic sequences, Proc. Computer Vision and Pattern Recogition (CVPR) Conf., pp. 2906-2912, (2009)

[8]

Micusik B., Kosecka J., Multi-view superpixel stereo in urban environments, Int. J. Comput. Vision, 89, pp. 106-119, (2010)

[9]

Gallup D., Frahm J.M., Mordohai P., Yang Q., Pollefeys M., Real-time plane-sweeping stereo with multiple sweeping directions, Proc. Computer Vision and Pattern Recogition, pp. 1-8, (2007)

[10]

Gallup D., Frahm J.M., Pollefeys M., Piecewise planar and non-planar stereo for urban scene reconstruction, Proc.Computer Vision and Pattern Recogition (CVPR) Conf., pp. 1418-1425, (2010)

← 1 2 →