Image Completion: Two-Staged Architecture

Liu, Hsin-Liang (2020) Image Completion: Two-Staged Architecture. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration manual) Download (2MB) | Preview

Abstract

Image completion or image inpainting refers to a technique that is to restore the images or remove unwanted objects and fill it with something that might have been there before. There are many methods for image inpainting for example the patch-based image completion. It searches the data and finds the patch that match the hole the best and complete the image with that patch. Nevertheless, it, sometimes, does work well and produce the unsatisfactory results. The patch can fit well in the hole for certain contents of the pictures but it fail once the contents are changed. In order to improve the performance and output the content-aware pictures, this research proposes a method that aims to provide more information about the missing pixels based on the symmetry of the face before it propagate to the neural network. Most of the current deep learning network take the damaged images and try to construct the new images without any knowledge about the missing portion of the images. Taking advantage of the face structures, we can improve the current algorithm so it get semantic image completion results. Our experiments show that it could lead to 5% improvements in term of SSIM and PSNR that measure the similarity of the two pictures. This research implements a dual-pipeline network structure(Zheng et al.; 2019) and this methodology can also be applied to other deep learning methods in an effort to improve the performance.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QA Mathematics > Computer software T Technology > T Technology (General) > Information Technology > Computer software
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Dan English
Date Deposited:	22 Jan 2021 15:26
Last Modified:	22 Jan 2021 15:26
URI:	https://norma.ncirl.ie/id/eprint/4454

Actions (login required)

View Item