A defect detection method based on wavelet domain signal catastrophe-points capture is proposed to address the difficulty of detecting minor defects on solar cell. The method, based on one-dimensional discrete signal, captures signal catastrophe-points through catastrophe-points detection on the image column by column in the wavelet domain, and corrects it with the energy centrobaric method so as to obtain the normal background of the defect image, which is finally fused with the original image by simple algebraic algorithm to highlight the defected area. Experiments show that the method is effective for detecting various defects on the surface of solar cells.