Digital Image Processing Using Sparse Representation Based on Iterative Methods

Please enable javascript in your browser.

Salemi, Gholamali | 2010

554 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 40515 (05)
University: Sharif University of Technology
Department: Electrical Engineering
Advisor(s): Marvasti, Farrokh
Abstract:
The main purpose of the thesis is Digital image Processing using the sparsity of image. The issue could be described in two ways. First is reconstruction of missed blocks of an image based on the sparsity of image in the transform domain and second is impulsive noise removal using the sparsity of noise in the spatial domain. In the first approach we will review Guleryuz method and simulate and analyze it. In the second approach, two new methods named RDE and Knockout will be introduced. These methods try two remove the impulsive noise of an image using its sparseness. RDE method is a development of conventional methods, but Knockout has a completely new idea. We will show that Knockout is able to completely remove the impulsive noise under specific conditions. Also we will see this method is independent of Image features and does not have the permanent image blurring problem of conventional methods. An upper and lower band will be introduced to guarantee the success of Knockout method and finally the performance of methods will be compared under these bands.
Keywords:
Image Processing ; Iteration Method ; Impulsive Noise Removal ; Sparse Signal Processing

No TOC