Automatic context-aware dust and scratch removal in scanned images

In this paper, we propose a novel method for automatic dust and scratch detection and removal in scanned images. The proposed method combines morphological operations with local image analysis that guides adaptive, image dependent perceptual defect thresholding. This allows us to achieve reliable separation of defects and intrinsic image features, and yields high quality image restoration. The method is capable of detecting and removing defects of very low contrast and of various sizes; it is efficient computationally and can be utilized in real time applications.