Objective detection of stain defects in low-light-level image intensifiers based on the discrete Fourier transform and edge feature analysis

Stain defects are one of the larger types of defects in low-light-level (LLL) image intensifiers. Their existence will weaken the target detection capability of the assembled equipment. The drawback of the traditional subjective detection method is the introduction of many artificial uncertainties. The algorithms of existing objective detection methods are relatively cumbersome and cannot be completely separated from human intervention. To address these issues, an objective detection method for stain defects based on the discrete Fourier transform and edge feature analysis is proposed. First, existing techniques are used to preprocess the target image to obtain a clean background. Then, the discrete Fourier transform and frequency bandpass filtering are utilized to extract image regions that may contain target defects. Afterward, a set of candidate image patches at adjacent scales is established, and their edge features are analyzed using the presented strategy to determine whether they belong to circular defects, linear defects, or non-defects. Finally, the dominant edges of the defective image patches are restored to the target image, thereby achieving objective detection of stain defects. To validate the performance of the proposed edge feature analysis strategy and defect detection method, traditional state-of-the-art related technologies are compared, and representative products are tested. The experimental results show that the presented edge feature analysis strategy can produce the most satisfactory results for the studied cases, and the proposed defect detection method outperforms existing objective detection techniques in terms of detection accuracy, with almost no need for human intervention. Therefore, it can be seen as a reasonable alternative to the testing method of existing stain defect detection systems for LLL image intensifiers.