A robust feature descriptor: Signed LBP

Chu Sing Yang, Yung Hsian Yang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We improve a texture descriptor, Local Binary Feature (LBP), called Signed Local Binary Pattern (SLBP) which is more robust in rotation and scale. In this paper we will introduce how to obtain more information in LBP, which with signed bit, and make feature more stable with mean of local area instead of single pixel’s intensity. Finally, to reach more robust in scale by difference smooth factor and implement by Integral Images to reduce computation cost. A pixel can perform different texture information in each scale, thus we select meaningful edge type in smallest scale. And signed bit is adding by current center area is greater or less then whole neighbor area. Then we implement cell and block concept from Histogram of Gradient to test the character recognition. In result part we prove SLBP have more robust than LBP in rotation, scale by texture image and natural scene image. The last part is testing the performance of recognition rate in IIIT5K database.

Original languageEnglish
Title of host publicationProceedings of the 2016 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2016
EditorsHamid R. Arabnia, Leonidas Deligiannidis, Fernando G. Tinetti, Jane You, George Jandieri, George Jandieri, Iakov Korovin, Gerald Schaefer, Kok Swee Sim, Ashu M. G. Solo
PublisherCSREA Press
Pages316-322
Number of pages7
ISBN (Electronic)1601324421, 9781601324429
Publication statusPublished - 2016 Jan 1
Event2016 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2016 - Las Vegas, United States
Duration: 2016 Jul 252016 Jul 28

Publication series

NameProceedings of the 2016 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2016

Conference

Conference2016 International Conference on Image Processing, Computer Vision, and Pattern Recognition, IPCV 2016
CountryUnited States
CityLas Vegas
Period16-07-2516-07-28

All Science Journal Classification (ASJC) codes

  • Computer Vision and Pattern Recognition
  • Computer Graphics and Computer-Aided Design

Fingerprint Dive into the research topics of 'A robust feature descriptor: Signed LBP'. Together they form a unique fingerprint.

Cite this