Autoencoder-based image dimensionality reduction methods

In this thesis, we study how images can be represented in a more compact way that still captures their most important features and preserves the similarities and dissimilarities between the images. These compact representations of images, also known as ‘image encodings’, allow us to identify similar images and image patches -- an operation very important for image processing and computer vision. By identifying similar image patches, we can perform operations such as image denoising, image inpainting, object tracking between frames in a video, and panorama image stitching. By identifying similar images, we can quickly retrieve images similar to a query image, for example, like in Google’s “search by image” feature. Throughout this thesis, we use a machine-learning--based method called autoencoder for learning these compact representations of images.

Grantor dc:publisher

Ghent University. Faculty of Engineering and Architecture

Year dc:date

2022

Author dc:creator

• Žižakić, Nina

Contributors dc:contributor

• Pizurica, Aleksandra

• Technology and Engineering

Statement dc:rights

• info:eu-repo/semantics/openAccess

Language dc:language

eng

Identifier

https://biblio.ugent.be/publication/01GJAPXMHGCPD9G2X30WR4MVH0
urn:isbn:9789463556422
https://biblio.ugent.be/publication/01GJAPXMHGCPD9G2X30WR4MVH0/file/01GJAR1SF8H8G34TAPX0KEF575

OAI identifier oai:identifier

oai:archive.ugent.be:01GJAPXMHGCPD9G2X30WR4MVH0