Publication detail
GP-GPU Implementation of the "Local Rank Differences" Image Feature
HEROUT, A. JOŠTH, R. ZEMČÍK, P. HRADIŠ, M.
Original Title
GP-GPU Implementation of the "Local Rank Differences" Image Feature
English Title
GP-GPU Implementation of the "Local Rank Differences" Image Feature
Type
conference paper
Language
en
Original Abstract
A currently popular trend in object detection and pattern recognition is usage of statistical classifiers, namely AdaBoost and its modifications. The speed performance of these classifiers largely depends on the low level image features they are using: both on the amount of information the feature provides and the processor time of its evaluation. Local Rank Di erences is an image feature that is alternative to commonly used haar wavelets. It is suitable for implementation in programmable (FPGA) or specialized (ASIC) hardware, but - as this paper shows - it performs very well on graphics hardware (GPU) used in general purpose manner (GPGPU, namely CUDA in this case) as well. The paper discusses the LRD features and their properties, describes an experimental implementation of the LRD in graphics hardware using CUDA, presents its empirical performance measures compared to alternative approaches, suggests several notes on practical usage of LRD and proposes directions for future work.
English abstract
A currently popular trend in object detection and pattern recognition is usage of statistical classifiers, namely AdaBoost and its modifications. The speed performance of these classifiers largely depends on the low level image features they are using: both on the amount of information the feature provides and the processor time of its evaluation. Local Rank Di erences is an image feature that is alternative to commonly used haar wavelets. It is suitable for implementation in programmable (FPGA) or specialized (ASIC) hardware, but - as this paper shows - it performs very well on graphics hardware (GPU) used in general purpose manner (GPGPU, namely CUDA in this case) as well. The paper discusses the LRD features and their properties, describes an experimental implementation of the LRD in graphics hardware using CUDA, presents its empirical performance measures compared to alternative approaches, suggests several notes on practical usage of LRD and proposes directions for future work.
Keywords
pattern recognition, adaptive boosting, AdaBoost, WaldBoost, image features, LRD, Local Rank Differences, hardware acceleration
RIV year
2008
Released
12.11.2008
Publisher
Springer Verlag
Location
Heidelberg
ISBN
978-3-642-02344-6
Book
Proceedings of International Conference on Computer Vision and Graphics 2008
Edition
Lecture Notes in Computer Science
Edition number
NEUVEDEN
Pages from
1
Pages to
11
Pages count
11
URL
Documents
BibTex
@inproceedings{BUT29556,
author="Adam {Herout} and Radovan {Jošth} and Pavel {Zemčík} and Michal {Hradiš}",
title="GP-GPU Implementation of the "Local Rank Differences" Image Feature",
annote="A currently popular trend in object detection and pattern
recognition is usage of statistical classifiers, namely AdaBoost and its
modifications. The speed performance of these classifiers largely depends on the
low level image features they are using: both on the amount of information the
feature provides and the processor time of its evaluation. Local Rank Di erences
is an image feature that is alternative to commonly used haar wavelets. It is
suitable for implementation in programmable (FPGA) or specialized (ASIC)
hardware, but - as this paper shows - it performs very well on graphics hardware
(GPU) used in general purpose manner (GPGPU, namely CUDA in this case) as well.
The paper discusses the LRD features and their properties, describes an
experimental implementation of the LRD in graphics hardware using CUDA, presents
its empirical performance measures compared to alternative approaches, suggests
several notes on practical usage of LRD and proposes directions for future work.",
address="Springer Verlag",
booktitle="Proceedings of International Conference on Computer Vision and Graphics 2008",
chapter="29556",
edition="Lecture Notes in Computer Science",
howpublished="print",
institution="Springer Verlag",
year="2008",
month="november",
pages="1--11",
publisher="Springer Verlag",
type="conference paper"
}