Hep-2 staining pattern classification is used for immune disease diagnosis.Due to the influence of imaging environment and the visual characteristics of cell images

the recognition accuracy of reported methods are still unsatisfied.A novel feature suitable for HEp-2 staining pattern classification is proposed.After constructing serial smoothing images with Gaussian scale parameter

the description ability of shape index is used to describe second-order image structure

moreover

the space structure can be presented by multi-threshold image segmentation.The proposed approach was tested in ICPR and SNPHEp-2 datasets by 2-flod cross validation in cell level

and achieved 89.83% and 87.49% accuracy for the two datasets respectively.Moreover

in 28 fold cross validation test of ICPR

the method achieved 60.5% and 70.56% accuracy in cell level and image level respectively.Experimental results show that the proposed method is superior to other popular texture descriptors

such as LBP and CLBP

and approximate to the performance of CoALBP feature.