Archive film defect detection and removal: an automatic restoration framework

Area of Science:

• Digital image processing
• Computer vision
• Film restoration

Background:

• Digitized archive films often suffer from defects like dirt and blotches.
• Manual restoration is time-consuming and labor-intensive.
• Automated solutions are needed for efficient film preservation.

Purpose of the Study:

• To develop an automatic system for detecting and removing defects in digitized archive films.
• To improve the accuracy and efficiency of film restoration processes.
• To compare the proposed system against existing state-of-the-art methods.

Main Methods:

• A two-module system: defect detection and defect removal.
• Defect detection combines temporal and spatial information using a Hidden Markov Model (HMM).
• Defect removal employs a multiscale framework with random walk-based exemplar searching and feature updating.

Main Results:

• The system accurately detects defective pixels using HMM and refines defect maps.
• Restoration involves multiscale replacement and feature updating for degraded pixels.
• The proposed system demonstrates improved accuracy in both detection and restoration compared to state-of-the-art methods.

Conclusions:

• The developed automatic system effectively restores digitized archive films by addressing dirt and blotches.
• The combination of HMM-based detection and multiscale restoration offers superior performance.
• This work contributes to advancing automated techniques for digital film preservation.