High contrast feature of flaws on curved smooth transparent surface wa s hard to be captured by available commercial lighting systems.In consideration of the limitation of illumination for surface inspection,a back-light system exploiting dynamic binary fringe was presented.The lighting projected a set of dynamic binary fringes with equal interval movement,which contributed the transition zone between bright and dark stripe t o cover the entire surface detected, and maximized the directional difference between light rays transmitted respecti vely by defect and flawless surface. The feature of defect with homogenous background could be obtained by averaging the curved transparent image sequences illuminated by dynamic binary fringe.Combined with edge extraction ba sed on wavelet transform and global threshold segmentation based on histogram distribution,the appearance de fects on both curved and flat surface were greatly improved.The test results of three dimensional curved mobile glass demonstrated that the method mentioned abo ve could successfully extract low contrast defects on transparent curved surface ,provide a reliable guarantee to recognize defects accurately,and has a good app lication prospect.