Co-segmentation Guided Hough Transform for Robust Feature Matching
1 National Taiwan University
2 Academia Sinica
In IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI 2015)
Feature matching by our approach. (a) Initial candidate graph is constructed. In this toy example, we find 2 candidates for each feature. (b) Each correspondence candidate is projected into homography space. Green circles represent correct candidate correspondences (green lines in(a)). The red cross represents one incorrect correspondence candidate for feature point P. (c) Correspondences selection after Hough voting. For point P, it still get the incorrect match since none of its candidates is correct. (d) Grouping result from co-segmentation. (e) For P, its correct correspondence candidate (blue square) is recommended by features grouped in (d). (f) After Co-Segmentation guided Hough Voting, the correct correspondence (blue line) is established for feature P.
We present an algorithm that integrates image co-segmentation into feature matching, and can robustly yield accurate and dense feature correspondences. Inspired by the fact that correct feature correspondences on the same object typically have coherent transformations, we cast the task of feature matching as a density estimation problem in the homography space. Specifically, we project the homographies of correspondence candidates into the parametric Hough space, in which geometric verification of correspondences can be activated by voting. The precision of matching is then boosted. On the other hand, we leverage image co-segmentation, which discovers object boundaries, to determine relevant voters and speed up Hough voting. In addition, correspondence enrichment can be achieved by inferring the concerted homographies that are propagated between the features within the same segments. The recall is hence increased. In our approach, feature matching and image co-segmentation are tightly coupled. Through an iterative optimization process, more and more correct correspondences are detected owing to object boundaries revealed by co-segmentation. The proposed approach is comprehensively evaluated. Promising experimental results on four datasets manifest its effectiveness.
technical paper
presentation slide

@article{ ChenLC:2015:cohv,
     author = {Hsin-I Chen and Yen-Yu Lin and Bing-Yu Chen},
     journal = {PAMI},
     title = {Co-segmentation Guided Hough Transform for Robust Feature Matching},
     year = {2015},
     volume = {37},
     number = {12},
     pages = {2388-2401},
1504 visitors since August 2013