• Contact
• SMPTE.org

• About
• Attend
• Exhibit
• Schedule
• Sessions
• Sponsors
• Media
• Partners

Exhibit Hall Open

Grand Ballroom Salons 3-6

Tuesday 1:00 pm- 7:00 pm
Wednesday 10:00 am- 5:00 pm
Thursday 10:00 am- 2:00 pm

Automatic Flicker Removal by Digital Restoration Techniques

Zhiyong Li (China Research Institute of Film Science & Technology, P.R. China)

To rescue the precious archival films which were on the brink of destruction for poor preservation conditions, China Archival Film Digital Restoration Project was approved in 2004 and started up from 2005. An investment of 280 million Yuan was financed for the first stage development with the object to restore 5000 domestic films in the first five years. To carry out the project, the automatic restoration methods are now urgently needed than ever. The aim of this paper is to report our recent related work. At first, flicker in Chinese archival films was analyzed and classified into two categories: global flicker and local flicker. On the basis of the survey for the algorithms reported in literature, two techniques were devised for two kinds of flicker, respectively. In virtue of its uniformity in time domain and spatial domain, the reduction of global flicker was comparatively easy. A simplified linear model was used and the parameters of the model could be estimated via the least mean square technique. The method is examined to be efficient and noise-tolerant by the massive experiments. Then we focused our main concerns on the treatment of local flicker. Considering its smoothness in time domain and spatial domain, random changes in position and along time line, the description and modeling of local flicker are very complicated. In terms of Tikhonov’s regularization theory, a cost functional was established by introducing a penalty function which reflects the physical characters of local flicker. To minimize the cost functional, the technique of partial differential equations (PDF) was utilized. The computational experiments show that the method works well. The algorithm converges quickly and it has a good adaptability for different shots of films. The processing is fully automatic. This is important for the reduction of man-power cost in massive processing application.

Thursday, Oct 29
8:30 AM - 11:00 AM

Sponsors