Quality Enhancement and Watermarking for Historical Images

This project, which brings together the expertise of the AImageLab research team and the industry experience of Memooria s.r.l., aims to develop cutting-edge technologies and techniques for digital watermark and machine learning tools for enhancing image quality and resolution.

The project has two main objectives:

• The primary objective is to develop prototypes for a print-resistant invisible digital watermark. To this aim, we will apply and develop innovative digital watermarking techniques that are both invisible to the human eye and resistant to printing and scanning processes. We will also conduct a comprehensive study of existing digital watermarking techniques and evaluate their effectiveness and resistance to the print-scan process in document applications and on very high resolution images. We will also develop new digital watermarking prototypes by combining classical techniques with low computational cost (such as frequency analysis or dither masks) and the use of convolutional neural networks. This approach will ensure that our digital watermarks are not only imperceptible to the human eye but also resistant to attempts at removal or alteration. The development of such a watermarking technique is crucial to ensure that historical digital images are protected from any attempts at counterfeiting or forgery.
   
• The second objective of this project is to develop machine learning architectures for enhancing image quality and resolution. The activity will involve creating deep learning algorithms to improve the perceived resolution of images and enhance their overall perceptual quality. This will be accomplished by implementing state-of-the-art super-resolution and image enhancement algorithms based on convolutional neural networks and new self-attentive architectures. Our results will enable an improvement in the the visual quality of images and make them more suitable for study and analysis, which is vital for the field of cultural heritage preservation.
   

Memooria will make available datasets of historical images and other digital material for the development and validation of the prototypes. Additionally, we will also use publicly available datasets present in the literature for both development and experimental testing.

The project is part of the "Tech Fast Lombardia" call, of which Memooria was the winner, promoted under Action I.1.b.1.2 of the POR FESR 2014-2020 of the Lombardy Region, with the project entitled "Opensource platform and advanced digitization and monitoring services for the artistic, archival and archaeological heritage. Storage services and distribution of high quality digital content via cloud platform”.

The preservation of cultural heritage is of paramount importance, and Memooria is dedicated to promoting the adoption of preventive conservation strategies. Their approach is based on the principle of minimum intervention, which means they aim to minimize the need for emergency restorations and instead prioritize research and study. To achieve this goal, Memooria utilizes 2D and 3D digitization, advanced analysis techniques, and secure storage of digital images. These tecniques allow for the objective detection and measurement of decay patterns, which can be detected earlier than with human perception alone. This approach is in line with digital preservation policies, and helps to ensure that cultural heritage is protected for future generations.

The collaboration between AImageLab and Memooria brings together the expertise of the AImageLab research team and the industry experience of Memooria, resulting in a unique and highly effective partnership. The project is expected to lead to advancements in the field of image processing, with a significant impact on cultural heritage preservation.