Search

The local binary pattern (LBP) descriptor is widely used in texture analysis because of its computational simplicity and robustness to illumination changes. However, LBP has limitations to fully capture discriminative information since only the sign information of the difference vector in a local region is used. To enhance the performance of LBP, we propose a new descriptor for texture classification—feature based local binary pattern (FbLBP). In the proposed FbLBP, difference vector is decomposed into sign part and magnitude part, the sign part is described by conventional LBP, while the magnitude part is described by two features of the mean and the variance of the magnitude vector. The way we extract magnitude information in difference vector shows high complementarity to the sign part and less sensitive to illumination changes with a low dimensionality. Furthermore, an adaptive local threshold is used to convert these two features into binary codes. The proposed low dimensional FbLBP is very fast to construct and no parameters are required to tune for different kinds of databases. Experimental results on four representative texture databases of Outex, CUReT, UIUC, and XU_HR show that the proposed FbLBP achieves more than 10% improvement compared with conventional LBP and 1%–3% improvement compared with the best classification accuracy among other benchmarked state-of-the-art LBP variants. [ABSTRACT FROM AUTHOR]

Local binary pattern (LBP) is sensitive to the noise and suffers from limited discriminative capability, and many LBP variants are reported in the recent literatures. Although a lot of significant progresses have been made, most LBP variants still have limitations of noise sensitivity, high dimensionality, and computational inefficiency. In view of this, we propose a new noise-robust local image descriptor named the diamond sampling structure-based local adaptive binary pattern (DLABP) in this letter, which aims at achieving both efficiency and simplicity at the same time. It mainly features three contributions: 1) an effective diamond sampling structure to decrease the feature dimensionality significantly by fixing the number of sampling neighbors to a constant of 8; 2) a simple and new “average method on the radial direction” to enhance the noise robustness; and 3) an effective adaptive quantization threshold strategy to restore the noise-corrupted nonuniform patterns back to possible uniform patterns. Extensive experiments are conducted on three benchmark texture databases of Outex, UIUC, and CUReT. Compared to state-of-the-art LBP-like methods, the proposed approach consistently demonstrates superior performances both in noise-free conditions and in the presence of high levels of noise, while it has a low complexity and a smaller feature dimension. [ABSTRACT FROM PUBLISHER]

Highlights • The paper presents a novel central pixel selection (CPS) strategy for LBP framework. • The proposed strategy can be applied to LBP and its variants. • The results of proposed CPS strategy are significant. Abstract Local binary pattern (LBP) has been successfully used in computer vision and pattern recognition applications, such as biomedical image analysis, remote sensing and image retrieval. However, the current LBP-based features, which assign a fixed sampling radius for all pixels in a single scale, completely ignore the fact that different central pixels actually have different local gray-value distributions and the proper sampling radius should be different for pixels. In this paper, we propose a novel and effective central pixel selection (CPS) strategy by using gradient information to classify central pixels of a texture image into different classes based on their local gray-value distributions. Then, we introduce this CPS strategy into the LBP framework and assign an adaptive sampling radius for each central pixel according to the class it belongs to. As a preprocessing step of LBP framework, this CPS strategy can also be integrated into any other LBP variants so as to extract more effective local texture features. Extensive experiments on five representative texture databases of Outex, UIUC, CUReT, UMD and ALOT validate the efficiency of the proposed central pixel selection (CPS) strategy, which can achieve almost 16% improvement over the original LBP and 1%–10% improvement compared with the best classification accuracy among other benchmarked state-of-the-art LBP variants. [ABSTRACT FROM AUTHOR]