Surface defects and texture features have a significant effect on the opticalcal properties of advanced optical components. However, most precision optical components have non-stochastic surfaces, so the current defect identification algorithm needs to be further improved to meet the quality inspection requirements for non-stochastic surfaces. In this paper, the scheme of non-subsampled contourlet transform is applied to identify feature of non-stochastic surfaces. A concrete analysis of sparse representation and feature identification about the non-subsampled contourlet transform were presented. The effectiveness of the method is proved by simulation results and experimental examples.