Abstract
dc:descriptionManually selecting photographs from large image collections is a time-consuming task, par- ticularly when photos are captured in quick succession and differ only in minor details such as sharpness, facial expression, or eye state. Existing solutions often require users to combine multi- ple tools for similarity detection, quality assessment, and facial analysis, which increases workflow complexity. This bachelor’s thesis presents the Smart Photo Culling Assistant, a web-based system designed to support and partially automate the photo selection process while preserving full user control over final decisions. The system integrates visual similarity detection, objective image quality assessment, and face and eye-state analysis into a single processing pipeline. Several image similarity algorithms, including SSIM, dHash, and pHash, were analysed and experimentally evaluated using a real-world photo dataset containing natural variations such as changes in facial expression and eye state. Eye-state detection was implemented using facial land- marks and the Eye Aspect Ratio (EAR) method, enabling reliable identification of images affected by unintended blinks. The experimental results demonstrate that perceptual hashing based on dHash provides the best balance between computational efficiency and practical applicability for photo culling tasks. Overall, the proposed system reduces manual effort during photo selection and delivers transparent, structured recommendations that assist users without replacing human judgmen.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Rudenko, Valeriia,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD277690199&prefLang=en_US