This study proposed a defect detection method based on the reflection moiré images to efficiently detect defects on the polished surface of a workpiece. The proposed method located defects by detecting variations in the moiré image reflected by the polished surface. The illumination model of the polished surface was analyzed, and the SHEN-Castan algorithm was used to suppress the edge-step effect of the moiré. Defects were detected and located after performing defect extraction to remove false defect via Gabor transform and maximum entropy segmentation. Experimental results and statistics show that the proposed method can detect surface defects on different polished workpiece surfaces at a detection rate of >92%. The proposed method can be used to independently set the detection resolution ratio of a system, thereby increasing the system’s adaptability to detect different types of defects and considerably improving its extensibility, universality, and practicability. Thus, the proposed method can efficiently detect defects on highly reflective polished workpiece surfaces and has great theoretical as well as economic value.