Surface defects are frequently observed in calendered polyvinyl chloride (PVC) films. Their evaluation in production environments is typically qualitative and operator dependent. Using the common gas entrapment defects as the test case, the present study develops a four-step image-processing workflow that converts scanned film images into pixel intensity matrices and groups defect pixels using density-based clustering (DBSCAN). The procedure provides quantitative measures of defect count, size, and spatial distribution without manual labeling. The effects of roll gap, calendering speed, upstream mixing time, and plasticizer type were examined under controlled conditions. Larger roll gaps and higher speeds reduced degassing efficiency and increased both defect number and defect area. Short mixing times led to incomplete gelation and higher defect frequency. Among the tested plasticizers, TOTM produced the lowest defect counts, followed by DEHP and ESBO. Design-of-experiments analysis ranked parameter sensitivity and identified operating ranges that limit defect formation. The method provides a practical basis for routine surface inspection and supports process adjustment using measurable defect metrics rather than visual judgment alone.