Year: 2024
Type: conference-paper
Abstract: Restoration of old photographs is a demanding task that requires a lot of skill and attention to detail. It can involve many different steps in photo editing and, until recently, was almost always done by photo editing professionals. Thanks to numerous photo editing AI applications, nowadays even the inexperienced users can perform this task. However, the results may not always be satisfactory. In this work, we evaluated the effectiveness of different free web-based AI applications in restoring old photographs that exhibited the most frequently seen damage. Tested AI solutions enabled significantly faster photo editing compared to manual techniques, while still providing visually pleasing results. On the other hand, they showed tendency to smooth out graininess, important details, and textures in images. When automatic processing failed to correct certain details, the results heavily depended on the user's skill to recognize and mark the errors for post-processing. Therefore, we believe that AI tools for photo restoration are a great choice for fast image processing, but their outcomes should be manually adjusted to ensure the authenticity and naturalness of old photographs.
Authors: ", Ivana Tomić, Nataša Bajšanski, Jelena Vladušić, Ivana Jurič
Institution: University of Novi Sad
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4403644071