Historical films usually have defects. We study the type of defects, and propose a series of solutions to detect defects before they are repaired by our inpainting algorithms. This paper focuses on a difficult issue to locate long vertical line defects in aged films. A progressive detection algorithm is proposed. We are able to detect more than 86% (recall rate) of effective line defects. These line defects are then removed step by step. The experiments use real historical video collected from national museum and public channel, instead of using computer generated noise. The results are visually pleasant based on our subjective evaluation by volunteers.
Timothy K. Shih, Louis H. Lin, Wonjun Lee