Statistical binary patterns for rotational invariant texture classification

摘要: A new texture representation framework called statistical binary patterns (SBPs) is presented. It consists in applying rotation invariant local pattern operators ( LBP riu 2 ) to a series of moment images, defined by statistics uniformly computed using given spatial support. can be seen as generalisation the commonly used complementation approach (CLBP), since it extends description not only contrast information, but also higher order variations. In short, SBPs aim at expanding self-similarity operator from grey level regional distribution level. Thanks richer description, have better discrimination power than other variants. Furthermore, thanks regularisation effect moments, SBP descriptors show noise robustness classical CLBPs. The interest validated through large experimental study performed on five databases: KTH-TIPS, KTH-TIPS 2b, CUReT, UIUC and DTD. results that, for four first datasets, are comparable or outperform recent state-of-the-art methods, even small support operator, limited size computation statistics. HighlightsWe extend pixel level.We exploit images calculated statistics.Statistical moments clearly improve expressiveness descriptor.