Spatio-temporal blotches detection and removal in archive video

In this paper, we propose a digital restoration algorithm dealing with a one of the most common defect in archived video so-called blotches. TWo main modules compose the algorithm: blotches detection and removal. For the first module, we propose efficient blotches detection algorithm based on spatio-temporal information. This is done by using a temporal median filter applied on the adjacent frames and spatial segmentation of the current frame to obtain candidate blotches regions. Furthermore, SDIx detector joining consensus voting strategy criteria, is applied on only candidate regions to separate blotches regions from the others. In the second step, we propose a novel iterative algorithm dedicated to remove blotches. This method assumes that the adjacent regions have high similarity in terms motion information. Thus, at any pixel belonging to blotches could be associated to a pixel in the adjacent frames by motion-compensation and using the motion vector of the neighboring regions. The proposed restoration system is compared against the state-of-the-art methods to demonstrate the system performance in detection and restoration stages using synthetic and real degraded image sequences.