O3S-MTP: Oriented star sampling structure based multi-scale ternary pattern for texture classification

This paper presents an effective hand-crafted texture operator for texture recognition, referred to as oriented star sampling structure based multi-scale ternary pattern (O3S-MTP) which is expected to better represent salient local texture structure. Instead of heuristic code constructions, the proposed O3S-MTP descriptor is defined as a gray-scale invariant texture measure, derived from a general definition of texture in the neighborhood of a 3 × 3 square patch thanks to the flexibility of the graph theory and dominating set. Different from the existing non-oriented local graph structure based texture descriptors, which cannot fully represent the extent of pixel difference, the essence of O3S-MTP operator is to encode the structure of local neighborhood by analyzing the differential excitation and orientation information using new effective oriented star sampling local structures. Extensive experiments on twelve challenging representative widely-used texture datasets show that O3S-MTP can achieve competitive and superior texture classification performance than a large number of recent most promising state-of-the-art texture descriptors. Furthermore, the obtained classification results were statistically validated through the Wilcoxon signed rank test based ranking method. •Based on the dominating set and graph theory, we introduce a new orientation graphic structure, which will contribute to solve the non-orientation problem of LGS (local graph structure) based methods (LGS, ELGS, SLGS, etc.).•We propose a new prominent image representation approach referred to as oriented star sampling structure based multi-scale ternary pattern (O3S-MTP) for texture classification.•Instead of binary coding, multi-level coding in different orientations is used as well.•Extensive evaluation on twelve challenging representative texture datasets is performed, showing that the proposed descriptor demonstrates superior performance to a large number of old and recent state-of-the-art LBP variants and non-LBP methods.•The optimized user-specified parameters for the parametric methods on each dataset is generated to represent and demonstrate the effectiveness and the stability of the proposed O3S-MTP operator.