Rotation invariant texture descriptors based on Gaussian Markov random fields for classification

•Rotation invariant descriptors based on Local Parameter Histograms are proposed.•Two approaches are suggested which produce RI-LPH and I-LPH descriptors.•RI-LPH descriptor is formulated by circular shifting the neighbour values.•I-LPH descriptor is generated using isotropic Gaussian–Markov Random Fields.•Both descriptors achieve higher classification accuracies in invariant texture analysis.