Towards automatic blotch detection for film restoration by comparison of spatio-temporal neighbours

  • Authors:
  • Peter Gaughran;Susan Bergin;Ronan Reilly

  • Affiliations:
  • Department of Computer Science, National University of Ireland, Maynooth, Co. Kildare, Ireland;Department of Computer Science, National University of Ireland, Maynooth, Co. Kildare, Ireland;Department of Computer Science, National University of Ireland, Maynooth, Co. Kildare, Ireland

  • Venue:
  • AICS'09 Proceedings of the 20th Irish conference on Artificial intelligence and cognitive science
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, a new method of blotch detection for digitised film sequences is proposed. Due to the aging of film stocks, their poor storage and/or repeated viewing, it is estimated that approximately 50% of all films produced prior to 1950 have either been destroyed or rendered unwatchable [1,2]. To prevent their complete destruction, original film reels must be scanned into digital format; however, any defects such as blotches will be retained. By combining a variation of a linear time, contour tracing technique with a simple temporal nearest neighbour algorithm, a preliminary detection system has been created. Using component labelling of dirt and sparkle the overall performance of the completed system, in terms of time and accuracy, will compare favourably to traditional motion compensated detection methods. This small study (based on 13 film sequences) represents a significant first step towards automatic blotch detection.