Robust Stereo Matching Using Probabilistic Laplacian Surface Propagation
This paper describes a probabilistic Laplacian surface propagation (PLSP) framework for a robust stereo matching under severe radiometric variations. We discover that a progressive scheme overcomes an inherent limitation for this task, while most conventi
- PDF / 1,910,695 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 40 Downloads / 208 Views
Yonsei University, Seoul, Republic of Korea {srkim89,ryus01,seonjookim,khsohn}@yonsei.ac.kr 2 Inria, Paris, France [email protected]
Abstract. This paper describes a probabilistic Laplacian surface propagation (PLSP) framework for a robust stereo matching under severe radiometric variations. We discover that a progressive scheme overcomes an inherent limitation for this task, while most conventional efforts have been focusing on designing a robust cost function. We propose the ground control surfaces (GCSs) designed as progressive unit, which alleviates the problems of conventional progressive methods and superpixel based methods, simultaneously. Moreover, we introduce a novel confidence measure for stereo pairs taken under radiometric variations based on the probability of correspondences. Specifically, the PLSP estimates the GCSs from initial sparse disparity maps using a weighted least-square. The GCSs are then propagated on a superpixel graph with a surface confidence weighting. Experimental results show that the PLSP outperforms state-of-the-art robust cost function based methods and other propagation methods for the stereo matching under radiometric variations.
1
Introduction
Stereo matching aims to extract 3D scene information by finding the correspondence between stereo pairs taken at different viewpoints of the same scene [1]. Nowadays, state-of-the-art methods provide satisfactory results under the color consistency condition, i.e., corresponding pixels have a similar color distribution. However, the color consistency assumption is often violated since the color of an image is the result of complex combinations of imaging pipelines. Specifically, various factors including illumination source variations, non-Lambertian surfaces, vignetting, device characteristics, and an image noise have an influence on the performance of the stereo matching [2]. Conventionally, to alleviate these problems, a number of methods have been proposed to develop a robust cost function that is insensitive to radiometric distortions [2–7]. However, for stereo images taken under challenging environments, e.g., severe radiometric variations, some pixels or regions cause erroneous local minima. In this case, a robust cost Bumsub Ham—WILLOW project-team, D´epartement d’Informatique de l’Ecole Normale Sup´erieure, ENS/Inria/CNRS UMR 8548. c Springer International Publishing Switzerland 2015 D. Cremers et al. (Eds.): ACCV 2014, Part I, LNCS 9003, pp. 368–383, 2015. DOI: 10.1007/978-3-319-16865-4 24
Robust Stereo Matching Using Probabilistic Laplacian Surface Propagation
369
function approach cannot guarantee to estimate reliable correspondences. In addition, costly global optimizations on the Markov random field (MRF), including a graph-cut (GC) and a belief propagation (BP) [8], cannot also infer a fully reliable solution and even propagate errors under these circumstances. We discover that a progressive framework can overcome such an inherent limitation for the stereo matching under severe radiometric variations. It is inspired by
Data Loading...
