Robust Detection of Single-frame Defects in Archived Film
Publication from Digital
Proceedings of International Conference on Pattern Recognition, 2012
The main issue in current algorithms for the detection of single-frame defects like dust, dirt and blotches in archived film is the significant number of false alarms due to motion compensation errors and film grain. This typically leads to disturbing artifacts occurring in the subsequent defect removal process. We propose a novel algorithm for the detection of single-frame defects which addresses this issue. A continuous response map is defined which indicates the probability of a pixel being part of a single-frame defect. Furthermore, we introduce the novel co-support which is applied to the response map for both noise suppression and spatial completion of potential defects. Finally, regions which are likely affected by motion compensation errors are
identified (e.g., by analyzing the motion field) and the singleframe-defect response map is damped in these regions. Experimental results show that the proposed method outperforms state-of-the-art algorithms in terms of accuracy and robustness to motion estimation issues.