A technique for the detection and removal of local defects in digital continuous-tone images.

摘要

The present invention is a method for automatically detecting and correcting a wide range of local digital image defects with minimal user intervention. The detection process employs brightness and color thresholds in conjunction with magnitude thresholds on residuals of nonlinear spatial filters to separate defects from scene content with minimal confusion. The detected defects are then cosmetically corrected by combinations of nonlinear smoothing and grey-scale erosion. Several options are outlined for the feature selection, detection, and cleaning operations depending on source type and computational constraints.