This study proposes a machine vision-based surface defect detection algorithm to enhance the effect and efficiency of detecting surface defects in polyvinyl chloride (PVC) pipes for industrial production. The algorithm performs image preprocessing and defect detection. Image preprocessing includes steps such as edge traversal, fringe detection, and Gamma transformation. Defect detection mainly includes horizontal and vertical projection, fast region growing for connected region marking, and block processing. The proposed algorithm accelerates the Gamma transformation and region growing, and it can also be used to optimally detect surface defects in PVC pipes, while avoiding false detection. Results of tests and actual factory inspections suggest that the proposed algorithm achieves a detection accuracy of 97.6%, with a real-time detection speed of >60 m/min, and a minimum defect detection area of 0.05 mm 2. Moreover, a unilateral jitter of <5 mm does not cause any false alarms and the missed detection rate is 0 when the pipe runs at a speed of 45 m/min, which meets actual production needs.