Please use this identifier to cite or link to this item: http://scholarbank.nus.edu.sg/handle/10635/47608
Title: DATA-DRIVEN FACIAL IMAGE SYNTHESIS FROM POOR QUALITY LOW RESOLUTION IMAGE
Authors: LOKE YUAN REN
Keywords: Image Superresolution, Image Inpaining, Markov Random Field, Facial Image Synthesis, iterative projection onto convex sets, PCA
Issue Date: 13-Feb-2012
Source: LOKE YUAN REN (2012-02-13). DATA-DRIVEN FACIAL IMAGE SYNTHESIS FROM POOR QUALITY LOW RESOLUTION IMAGE. ScholarBank@NUS Repository.
Abstract: In this thesis, we are interested in automatically restoring the missing information in the corrupted images and generating a visually-pleasing super-resolution facial image from a low resolution image with a set of high resolution training images. We propose a learned guidance vector field based on a Principal Component Analysis model for image inpainting, and the results are seamless and the structure of face is preserved. In image super-resolution (SR) on generic face, an iterative projection onto convex sets algorithm is proposed to optimize the model and data constraints. In image SR on specific face, an image retrieval system for pose and expression is proposed. The selected high resolution images are used as the candidates for the image SR, and a Markov Random Field model based on color and edge constraints is used to find the optimum solution. The proposed approach resolves the blur and noisy images, and significantly improves the performance.
URI: http://scholarbank.nus.edu.sg/handle/10635/47608
Appears in Collections:Ph.D Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
LOKEYR.pdf1.97 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

168
checked on Dec 11, 2017

Download(s)

41
checked on Dec 11, 2017

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.