Institutional Repository of Vilnius University
Nuotraukos spalvų korekcija naudojant vaizdų transformacijas /
Abstract
dc:descriptionExisting automatic exposure and white balance correction tools that do not use deep learning techniques perform exposure and white balance correction in an unstable and poor quality way. In this work, Pix2Pix GAN and InstructPix2Pix models are trained to perform automatic exposure and white balance correction. A loss function has also been developed for the exposure and white balance task and a neural network model has been trained using this loss function. To evaluate the results generated by these trained models, the averages of the pixel colour differences are computed using the CIEDE2000 formula and the SSIM indices. The results are compared between the trained models. The model with the best results is compared with the automatic exposure and white balance correction functionalities of Adobe Lightroom Classic. When comparing the models, the model trained with the developed loss function showed better and higher quality results than Pix2Pix GAN, InstructPix2Pix and Adobe Lightroom Classic autoexposure and white balance correction functionalities.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Serafimavičius, Redas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210647951&prefLang=en_US