Real-time texture extraction based on the improved median robust extended local binary pattern

Mengnan Wang, Yanjun Zhang*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

Local binary patterns (LBP) are considered as the most computation efficient and high-performance texture features. Among all variants of the LBPs, Median Robust Extended Local Binary Pattern (MRELBP) [1] is considered as one of the highest-performance one. Based on the MRELBP, this paper proposes an improved real-time texture extraction architecture implemented on a field-programmable gate array (FPGA) device. The fixed-point fraction is used to replace the double float-point fraction, only results in less than 1% difference to the accuracy. The proposed system is implemented on Xilinx Zynq-7045, and the extraction time is linear with the size of the input image. The extraction time of the 128x128 image in the dataset Outex-TC is merely 85.64μs, which is more than 4800 times faster than Liu et al. implemented on MATLAB [2].

Original languageEnglish
Title of host publicationICCPR 2020 - Proceedings of 2020 9th International Conference on Computing and Pattern Recognition
PublisherAssociation for Computing Machinery
Pages280-287
Number of pages8
ISBN (Electronic)9781450387835
DOIs
Publication statusPublished - 30 Oct 2020
Event9th International Conference on Computing and Pattern Recognition, ICCPR 2020 - Virtual, Online, China
Duration: 30 Oct 20201 Nov 2020

Publication series

NameACM International Conference Proceeding Series

Conference

Conference9th International Conference on Computing and Pattern Recognition, ICCPR 2020
Country/TerritoryChina
CityVirtual, Online
Period30/10/201/11/20

Keywords

  • Feature extraction
  • Field-Programmable Gate Array (FPGA)
  • MRELBP
  • Real-time

Fingerprint

Dive into the research topics of 'Real-time texture extraction based on the improved median robust extended local binary pattern'. Together they form a unique fingerprint.

Cite this