Please use this identifier to cite or link to this item:
|Title:||High quality image reconstruction from multiple low quality images||Authors:||Royyuru Mohana Prasanna Vadan||Keywords:||DRNTU::Engineering::Computer science and engineering||Issue Date:||2014||Abstract:||The purpose of this project was to implement an image reconstruction procedure from multiple low quality input images. It would integrate concepts of deblocking and deblurring for achieving a high performance in reconstruction. The main trends in this field were analyzed through the study of past projects and research undertaken by different universities and institutions. It was observed that a wide variety of concepts were available for use and implementation. The primary importance of this project was to integrate different theories in an optimal manner to achieve implementation of something existing in a new way. Reconstruction is based on enforcing the pixel consistency property. The relationship between a pixel and its neighborhood pixels between the multiple low quality images should be consistent for the reconstruction to be carried out. The reconstruction procedure follows the Piecewise Image Model, where the parameters of each pixel in the model are estimated from the different degraded input images. While the parameters of each pixel are being calculated, the pixels of the different input images are deblocked and deblurred. This generates the high quality reconstructed image. Post processing includes regularization of such parameters to within a reasonable range, followed by performance evaluation. Multiple images with different distortions of different levels were generated from a training image set. Experimental results on such distorted images have demonstrated that the reconstruction method can effectively generate a high quality image, and alleviate distortions in terms of blocking and blurring, while simultaneously preserving the detailed information. Better quality images in terms of both objective and subjective measurements were produced. The functionality of the project was discussed in parallel with existing methodologies and research works. The limitations of the implementation brought to light the recommendations for future development. The project hopes to contribute to the ever advancing research techniques of image reconstruction. It aims to integrate existing research works in a novel way and to generate results that could be further explored upon.||URI:||http://hdl.handle.net/10356/59008||Rights:||Nanyang Technological University||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||SCSE Student Reports (FYP/IA/PA/PI)|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.