Improving the Performance of Spiral Local Binary Pattern using Edge Information

Abstract

In many texture recognition problems, Local Binary Pattern (LBP) is used as texture descriptor and has achieved outstanding performances. Because of the success on the texture analysis, it is widely studied by many researchers and many variants of LBP are proposed. Edges of an image carry important information that can be used to extract discriminative features. In this paper, we investigate the importance of the information derived using Spiral Local Binary Pattern (SLBP) on the edges of a texture image. Also, we combine this information with SLBP features extracted from the whole image. Linear Regression Classification (LRC) is used for the classification process. Our experimental results in two face databases, namely CURet and UIUC texture databases, show that spiral topology-based LBP features extracted from the edges of an image has important discriminative information that improve classification accuracy.