Fast Approximate Nearest-Neighbor Field by Cascaded Spherical Hashing
Iban Torres (Technicolor)
Jordi Salvador (Technicolor)
Eduardo Pérez-Pellitero (Technicolor)
Proceedings of the Asian Conference on Computer Vision, 2014
The supplementary material can be accessed from here and the conference poster from here.
Abstract
We present an efficient and fast algorithm for computing approximate
nearest neighbor fields between two images. Our method builds on the
concept of Coherency-Sensitive Hashing (CSH), but uses a recent hashing
scheme, Spherical Hashing (SpH), which is known to be better adapted
to the nearest-neighbor problem for natural images. Cascaded Spherical
Hashing concatenates different configurations of SpH to build larger
Hash Tables with less elements in each bin to achieve higher selectivity.
Our method is able to amply outperform existing techniques like PatchMatch
and CSH. The parallelizable scheme has been straight-forwardly implemented
in OpenCL, and the experimental results show that our algorithm is
faster and more accurate than existing methods.
BibTeX
@inproceedings { Torres2014,author = {Torres, I. and Salvador, J. and P\'erez-Pellitero, E.},
title = {{Fast Approximate Nearest-Neighbor Field by Cascaded Spherical Hashing}},
booktitle = {Proc. Asian Conf. on Computer Vision},
pages = {first--last},
year = {2014},
}
