Categories: "Computer Science"
Reference #: 2012-010
Historical documents and images often become degraded over time due to factors such as aging and improper storage. Pictures, handwriting and even typewriter ink may become difficult to distinguish over time. Georgetown University has identified a novel image enhancement technique, which automatically transforms degraded images into legible images. This image enhancement invention analyzes pixel features of an image and can filter for only those pixels needed for document legibility. This invention is an improvement on prior image enhancement systems because it can enhance legibility of images effectively without the need of training data accumulated over time and it is not limited to a specific language or script.
In tests, this invention has been shown to significantly increase the legibility of documents such as aged handwritten letters, typed documents, and photographs. Specifically, documents that have suffered from fading or obscured by a darkening background can have the clarity of important features restored. Therefore, this technique for improving the legibility for documents will be of great use to such professionals as librarians and historians, and will aid efforts to preserve details from important manuscripts and records. It can also be incorporated into optical character recognition techniques for battlefield document recovery.
- Fully automated enhancement
- No training data required
- Language and script independent
U.S. Published Patent Application 20130129247
PCT Application Publication WO2013078182
Jon Parker, Gideon Frieder, Ophir Frieder