LTD: Local Ternary Descriptor for image matching

Author

Yongqiang Gao ; Yu Qiao ; Zhifeng Li ; Chunjing Xu

Author_Institution

Shenzhen Key Lab. of Comput. Vision & Pattern Recognition, Shenzhen Inst. of Adv. Technol., Chinese Univ. of Hong Kong, Shenzhen, China

fYear

2013

fDate

26-28 Aug. 2013

Firstpage

1375

Lastpage

1380

Abstract

Binary descriptors are receiving extensive research interests due to their storage and computation efficiency. A good binary descriptor should deliver sufficient information as well as be robust to image deformation and distortion. Recently, Calonder et al proposed Binary Robust Independent Elementary Features (BRIEF), which showed good performance in image matching. In this paper, we extend BRIEF to a Local Ternary Descriptor (LTD). Compared with BRIEF, LTD introduces a threshold to describe the difference of two pixels into three values. Our ternary descriptor can deliver more discriminative information than BRIEF while being robust to image deformation. We examine the key-point matching performance of LTD on several public datasets. The experimental results exhibit that LTD outperforms BRIEF.

Keywords

computational complexity; image matching; BRIEF; LTD; binary descriptors; binary robust independent elementary features; computation efficiency; discriminative information; image deformation; image distortion; image matching; key-point matching performance; local ternary descriptor; public datasets; Brightness; Feature extraction; Hamming distance; Histograms; Matched filters; Noise; Robustness; Hamming distance; binary pattern; descriptors; local ternary descriptor;

fLanguage

English

Publisher

ieee

Conference_Titel

Information and Automation (ICIA), 2013 IEEE International Conference on

Conference_Location

Yinchuan

Type

conf

DOI

10.1109/ICInfA.2013.6720508

Filename

6720508