Scale- and rotation-invariant texture description with improved local binary pattern features

• This paper presents a novel improved LBP-based descriptor for texture analysis.
• The proposed descriptor is adaptive to handle scaling and rotation changes.
• We proposed a technique to incorporate magnitude of the differences into LBP.
• Our method improves texture classification with respect to other LBP-based descriptors.

Local Binary Pattern (LBP) is an effective image descriptor based on joint distribution of signed gray level differences. Simplicity, discriminative power, computational efficiency and robustness to illumination changes are main properties of LBP. However, LBP is sensitive to scaling, rotation, viewpoint variations, and non-rigid deformations. In order to overcome these disadvantages of LBP, this paper proposes an improved LBP features. In our method, a circular neighboring radius and a dominant orientation are assigned to each pixel. To achieve scale invariance, we used the radius of blob-like structures to determine the circular neighboring set of each central pixel. Definition of LBP operator with respect to dominant orientation of each pixel can guarantee the rotation invariance of LBP features. Unlike original LBP operator which discards the magnitude information of the difference between the center and the neighbor gray values in a local neighborhood, a weighted LBP features is proposed in this paper. Several experiments are conducted to compare the proposed method with seven LBP-based descriptors for texture retrieval and classification using four databases: Brodatz, Outex, UIUC and UMD. Experimental results show that the proposed Weighted, Rotation- and Scale- Invariant Local Binary Pattern (WRSI_LBP) outperforms other LBP-based methods.