Tear Removal

author: David Corrigan
keywords: Tear Removal, Motion Estimation, Image Segmentation, Digital Movie Restoration

Tear Removal is concerned with the restoration of sequences in which the film material has been physically torn. In torn frames, two issues need to be addressed. Firstly, the regions divided by the tear need to be re-aligned. This can be achieved by estimating the motion of the two regions. Once the regions have been aligned correctly, the second stage of the restoration is to recover the damage image data along the tear boundary. A solution is found by treating it as a missing data problem. The data is interpolated from the appropriate regions in neighbouring frames.

Tear Removal must address the following challenges

  1. Detection of any torn frames in a sequence [1].
  2. Division of torn frames into regions divided by the tear boundary [2].
  3. Estimation of and compensation for the relative displacement between the regions [1].
  4. Recovery of damaged image data [3].

Examples

Figure 1: A torn frame
Figure 2: The segmented frame. The colours (red/green) represent the region to which each pixel is assigned
Figure 3: The segmented frame with displacement removed. Edges across tear are aligned.
Figure 4: The restored frame after damaged data is recovered.

Bibliography

  1. Automatic Treatment of Film Tear in Degraded Archived Media. David Corrigan and Anil Kokaram (2004) In ICIP '04. Singapore, October.
  2. Automated Segmentation of Torn Frames using the Graph Cuts Technique. David Corrigan, Naomi Harte and Anil Kokaram (2007) In IEEE International Conference on Image Processing (ICIP). San Antonio, Texas, USA, September. ( pdf )
  3. MCMC for joint noise reduction and missing data treatement in Degraded Video. A. Kokaram and S. J. Godsill (2002) IEEE Transactions on Signal Processing, Special Issue on MCMC, 50(2):189-205.
Page last modified on September 13, 2007