A Top-Down Approach to the Estimation of Depth Maps Driven by Morphological Segmentations

Abstract : Given a pair of stereo images, the spatial coordinates of a scene point can be derived from its projections onto the two considered image planes. Finding the correspondences between such projections however remains the main difficulty of the depth estimation problem: the matching of points across homogeneous regions is ambiguous and occluded points cannot be matched as their projections do not exist in one of the image planes. Instead of searching for dense point correspondences, this article proposes an approach to the estimation of depth map which is based on the matching of regions. The matchings are performed at two segmentation levels obtained by morphological criteria which ensure the existence of an hierarchy between the coarse and fine partitions. The hierarchy is then exploited in order to compute fine regional disparity maps which are accurate and free from noisy measurements. We finally show how this method fits to different sorts of stereo images: those which are highly textured, taken under constant illumination such as Middlebury and those which relevant information resides in the contours only.
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01158707
Contributor : Jean-Charles Bricola <>
Submitted on : Monday, June 1, 2015 - 5:38:09 PM
Last modification on : Monday, November 12, 2018 - 10:55:26 AM

Identifiers

Relations

Citation

Jean-Charles Bricola, Michel Bilodeau, Serge Beucher. A Top-Down Approach to the Estimation of Depth Maps Driven by Morphological Segmentations. International Symposium on Mathematical Morphology 2015, May 2015, Reykjavík, Iceland. Mathematical Morphology and Its Applications to Signal and Image Processing. 12th International Symposium, ISMM 2015, Reykjavik, Iceland, May 27-29, 2015. Proceedings, 9082, 2015, Lecture Notes in Computer Science. 〈http://link.springer.com/chapter/10.1007/978-3-319-18720-4_11〉. 〈10.1007/978-3-319-18720-4_11〉. 〈hal-01158707〉

Share

Metrics

Record views

302