Texture Representation via Joint Statistics of Local Quantized Patterns

Tiecheng SONG  Linfeng XU  Chao HUANG  Bing LUO  

Publication
IEICE TRANSACTIONS on Information and Systems   Vol.E97-D   No.1   pp.155-159
Publication Date: 2014/01/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E97.D.155
Print ISSN: 0916-8532
Type of Manuscript: LETTER
Category: Image Recognition, Computer Vision
Keyword: 
texture classification,  Local Binary Patterns (LBP),  Gaussian derivative filter,  ternary coding,  

Full Text: PDF(1.2MB)>>
Buy this Article




Summary: 
In this paper, a simple yet efficient texture representation is proposed for texture classification by exploring the joint statistics of local quantized patterns (jsLQP). In order to combine information of different domains, the Gaussian derivative filters are first employed to obtain the multi-scale gradient responses. Then, three feature maps are generated by encoding the local quantized binary and ternary patterns in the image space and the gradient space. Finally, these feature maps are hybridly encoded, and their joint histogram is used as the final texture representation. Extensive experiments demonstrate that the proposed method outperforms state-of-the-art LBP based and even learning based methods for texture classification.