Blotch detection in archived video based on regions matching

This paper turns on blotches detection in archived video. Blotches could be considered as the most common artifacts in archived video. This type of artifacts is not temporally correlated, it can be considered as impulse noise in time. To detect blotches, the proposed approach is performed into two mains steps: blotch regions candidate extraction and classification of these regions as blotches or not. The first step is performed by using sudden changes region detection by means of median filtering in time. The second step relies on finding similarity of the candidate region in the adjacent frames in gradient space. The proposed method has been compared against ROD, SROD and SDIx detectors. The obtained results show that our proposed approach outperforms the considered methods.