Learned Collaborative Stereo Refinement

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

In this work, we propose a learning-based method to denoise and refine disparity maps of a given stereo method. The proposed variational network arises naturally from unrolling the iterates of a proximal gradient method applied to a variational energy defined in a joint disparity, color, and confidence image space. Our method allows to learn a robust collaborative regularizer leveraging the joint statistics of the color image, the confidence map and the disparity map. Due to the variational structure of our method, the individual steps can be easily visualized, thus enabling interpretability of the method. We can therefore provide interesting insights into how our method refines and denoises disparity maps. The efficiency of our method is demonstrated by the publicly available stereo benchmarks Middlebury 2014 and Kitti 2015.
Original languageEnglish
Title of host publicationGerman Conference on Pattern Recognition
Pages3-17
Publication statusPublished - 2019

Fingerprint

Color
Gradient methods
Statistics

Cite this

Knöbelreiter, P., & Pock, T. (2019). Learned Collaborative Stereo Refinement. In German Conference on Pattern Recognition (pp. 3-17)

Learned Collaborative Stereo Refinement. / Knöbelreiter, Patrick; Pock, Thomas.

German Conference on Pattern Recognition. 2019. p. 3-17.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Knöbelreiter, P & Pock, T 2019, Learned Collaborative Stereo Refinement. in German Conference on Pattern Recognition. pp. 3-17.
Knöbelreiter P, Pock T. Learned Collaborative Stereo Refinement. In German Conference on Pattern Recognition. 2019. p. 3-17
Knöbelreiter, Patrick ; Pock, Thomas. / Learned Collaborative Stereo Refinement. German Conference on Pattern Recognition. 2019. pp. 3-17
@inproceedings{8d9794d4beb44fb592ede01f43cb874e,
title = "Learned Collaborative Stereo Refinement",
abstract = "In this work, we propose a learning-based method to denoise and refine disparity maps of a given stereo method. The proposed variational network arises naturally from unrolling the iterates of a proximal gradient method applied to a variational energy defined in a joint disparity, color, and confidence image space. Our method allows to learn a robust collaborative regularizer leveraging the joint statistics of the color image, the confidence map and the disparity map. Due to the variational structure of our method, the individual steps can be easily visualized, thus enabling interpretability of the method. We can therefore provide interesting insights into how our method refines and denoises disparity maps. The efficiency of our method is demonstrated by the publicly available stereo benchmarks Middlebury 2014 and Kitti 2015.",
author = "Patrick Kn{\"o}belreiter and Thomas Pock",
year = "2019",
language = "English",
pages = "3--17",
booktitle = "German Conference on Pattern Recognition",

}

TY - GEN

T1 - Learned Collaborative Stereo Refinement

AU - Knöbelreiter, Patrick

AU - Pock, Thomas

PY - 2019

Y1 - 2019

N2 - In this work, we propose a learning-based method to denoise and refine disparity maps of a given stereo method. The proposed variational network arises naturally from unrolling the iterates of a proximal gradient method applied to a variational energy defined in a joint disparity, color, and confidence image space. Our method allows to learn a robust collaborative regularizer leveraging the joint statistics of the color image, the confidence map and the disparity map. Due to the variational structure of our method, the individual steps can be easily visualized, thus enabling interpretability of the method. We can therefore provide interesting insights into how our method refines and denoises disparity maps. The efficiency of our method is demonstrated by the publicly available stereo benchmarks Middlebury 2014 and Kitti 2015.

AB - In this work, we propose a learning-based method to denoise and refine disparity maps of a given stereo method. The proposed variational network arises naturally from unrolling the iterates of a proximal gradient method applied to a variational energy defined in a joint disparity, color, and confidence image space. Our method allows to learn a robust collaborative regularizer leveraging the joint statistics of the color image, the confidence map and the disparity map. Due to the variational structure of our method, the individual steps can be easily visualized, thus enabling interpretability of the method. We can therefore provide interesting insights into how our method refines and denoises disparity maps. The efficiency of our method is demonstrated by the publicly available stereo benchmarks Middlebury 2014 and Kitti 2015.

M3 - Conference contribution

SP - 3

EP - 17

BT - German Conference on Pattern Recognition

ER -