Robust Local Graph Structure for Texture Classification

Local Binary Pattern (LBP) is one of the most popular operators in computer vision and digital image processing due to its simplicity and unique property of capturing local features. Yet, the local binary pattern operator has some drawbacks such as sensitivity to noise and illumination changes and this because of the thresholding process. Recently, Local Graph Structure (LGS) is proposed as an alternative solution to (LBP) to overcome the thresholding process by encoding the pattern based on the relationship of the pixels that form the local graph. In addition, (LGS) is investigated in many areas such as face recognition, spoofing and plant identification. Hence, in this paper we extend the idea of (LGS) and propose a new operator called Robust Local Graph Structure (RLGS) which utilizes the standard deviation of the local graph to capture more spatial information. Finally, the reported experiments results on (UIUC) texture database showed a significant increase in the performance with compare to other texture operators.