Fast Edge-Preserving PatchMatch for Large Displacement Optical Flow

Linchao Bao; Qingxiong Yang; Hailin Jin

doi:10.1109/CVPR.2014.452

Fast Edge-Preserving PatchMatch for Large Displacement Optical Flow

Linchao Bao, Qingxiong Yang, Hailin Jin

Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We present a fast optical flow algorithm that can handle large displacement motions. Our algorithm is inspired by recent successes of local methods in visual correspondence searching as well as approximate nearest neighbor field algorithms. The main novelty is a fast randomized edge-preserving approximate nearest neighbor field algorithm which propagates self-similarity patterns in addition to offsets. Experimental results on public optical flow benchmarks show that our method is significantly faster than state-of-the-art methods without compromising on quality, especially when scenes contain large motions.

Original language	English
Title of host publication	2014 IEEE Conference on Computer Vision and Pattern Recognition
Publisher	IEEE
Pages	3534-3541
Number of pages	8
ISBN (Print)	978-1-4799-5118-5
DOIs	https://doi.org/10.1109/CVPR.2014.452
Publication status	Published - 28 Jun 2014
Event	2014 IEEE Conference on Computer Vision and Pattern Recognition - Columbus, OH, USA Duration: 23 Jun 2014 → 28 Jun 2014

Conference

Conference	2014 IEEE Conference on Computer Vision and Pattern Recognition
Period	23/06/14 → 28/06/14

Keywords

Data structures
Boolean functions
Optical imaging
Approximation algorithms
Benchmark testing
Vectors
Estimation

Access to Document

10.1109/CVPR.2014.452

https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6909847

Cite this

@inproceedings{305479280fa24f86b42d6e2d3d2c290e,

title = "Fast Edge-Preserving PatchMatch for Large Displacement Optical Flow",

abstract = "We present a fast optical flow algorithm that can handle large displacement motions. Our algorithm is inspired by recent successes of local methods in visual correspondence searching as well as approximate nearest neighbor field algorithms. The main novelty is a fast randomized edge-preserving approximate nearest neighbor field algorithm which propagates self-similarity patterns in addition to offsets. Experimental results on public optical flow benchmarks show that our method is significantly faster than state-of-the-art methods without compromising on quality, especially when scenes contain large motions.",

keywords = "Data structures, Boolean functions, Optical imaging, Approximation algorithms, Benchmark testing, Vectors, Estimation",

author = "Linchao Bao and Qingxiong Yang and Hailin Jin",

year = "2014",

month = jun,

day = "28",

doi = "10.1109/CVPR.2014.452",

language = "English",

isbn = "978-1-4799-5118-5",

pages = "3534--3541",

booktitle = "2014 IEEE Conference on Computer Vision and Pattern Recognition",

publisher = "IEEE",

note = "2014 IEEE Conference on Computer Vision and Pattern Recognition ; Conference date: 23-06-2014 Through 28-06-2014",

}

TY - GEN

T1 - Fast Edge-Preserving PatchMatch for Large Displacement Optical Flow

AU - Bao, Linchao

AU - Yang, Qingxiong

AU - Jin, Hailin

PY - 2014/6/28

Y1 - 2014/6/28

N2 - We present a fast optical flow algorithm that can handle large displacement motions. Our algorithm is inspired by recent successes of local methods in visual correspondence searching as well as approximate nearest neighbor field algorithms. The main novelty is a fast randomized edge-preserving approximate nearest neighbor field algorithm which propagates self-similarity patterns in addition to offsets. Experimental results on public optical flow benchmarks show that our method is significantly faster than state-of-the-art methods without compromising on quality, especially when scenes contain large motions.

AB - We present a fast optical flow algorithm that can handle large displacement motions. Our algorithm is inspired by recent successes of local methods in visual correspondence searching as well as approximate nearest neighbor field algorithms. The main novelty is a fast randomized edge-preserving approximate nearest neighbor field algorithm which propagates self-similarity patterns in addition to offsets. Experimental results on public optical flow benchmarks show that our method is significantly faster than state-of-the-art methods without compromising on quality, especially when scenes contain large motions.

KW - Data structures

KW - Boolean functions

KW - Optical imaging

KW - Approximation algorithms

KW - Benchmark testing

KW - Vectors

KW - Estimation

UR - https://ieeexplore.ieee.org/document/6909847/

U2 - 10.1109/CVPR.2014.452

DO - 10.1109/CVPR.2014.452

M3 - Conference contribution

SN - 978-1-4799-5118-5

SP - 3534

EP - 3541

BT - 2014 IEEE Conference on Computer Vision and Pattern Recognition

PB - IEEE

T2 - 2014 IEEE Conference on Computer Vision and Pattern Recognition

Y2 - 23 June 2014 through 28 June 2014

ER -

Fast Edge-Preserving PatchMatch for Large Displacement Optical Flow

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this