An Image Quality Assessment Method for Surface Defect Inspection

The primary goal in developing an automatic defect inspection system is to obtain good-quality images. High image quality helps image detection methods extract defect features. Thus, this study proposes a comprehensive evaluation index to evaluate the image quality of image datasets for training defect detection models. Obtained images can be evaluated by the proposed index immediately as long as the lighting configuration is changed. The index consists of three parts: the image visibility, the dispersion of the image visibility of the dataset, and the image overexposure. Experiments validated that the comprehensive evaluation index was more consistent with the F2-score than the defect visibility using a YOLO defect detection model.