Please use this identifier to cite or link to this item: http://hdl.handle.net/10603/224664
Title: Mathematical Modeling and Simulation of Digital Image Processing Problems
Researcher: Thapliyal Chanda
Guide(s): Rana U.S.
Keywords: ART 1 neural network, artificial neural network, image deblurring, image data compression, pattern recognition, SNRout, single layer perceptron learning algorithm, plasticity, Huffman codes and B-codes.
Engineering and Technology,Computer Science,Computer Science Artificial Intelligence
University: Uttarakhand Technical University
Completed Date: 22-11-2017
Abstract: The objective of this thesis is to model and simulate different image processing problems. By identifying the crucial parameters in existing image processing techniques and in artificial neural network models, algorithms for image deblurring, image data compression and pattern recognition are modeled and simulated with better accuracies. newlineA general problem faced in captured image is due to the blur and it is also observed that some amount of blurring inevitably occurs in recording of digital images. Six deblurring algorithms have been proposed in this thesis. These algorithms are proposed alongwith the safety window. The performance of these algorithms for background removal is tested quantitatively by calculating SNRout. Pattern recognition is also an important task in image processing applications. Today we have devices for easier interface between human and computer. But it is necessary for more realistic human computer cooperation as computer usage reaches to a bigger mass. Character recognition is one such option. In this thesis, the perceptron learning rule is used for pattern recognition of noisy characters. In comparison to multilayer perceptron neural network this rule reduces the complexity of the network. newlineIn real time applications when environment is constantly changing, an autonomous learning system is needed, which maintains its plasticity or adaptability to the significant incidents and can also simultaneously stabilizes itself to the insignificant incidents. Adaptive resonance theory, ART provides a solution to this. The thesis also proposes the pattern recognition through ART1 neural network. In digital image processing applications for the sake of saving disk memory space and reducing transmission rate, images need to be compressed. For JPEG and JPEG 2000 images there are many techniques and standards for data compression. Huffman codes and B codes are normally used in the entropy coding phase. These compression techniques have been compared in this thesis for entropy and average word length. newline
Pagination: 148 pages
URI: http://hdl.handle.net/10603/224664
Appears in Departments:Department of Mathematics

Files in This Item:
File Description SizeFormat 
01-title page.pdfAttached File74.5 kBAdobe PDFView/Open
02-certificate.pdf232.48 kBAdobe PDFView/Open
03-contents.pdf1.37 MBAdobe PDFView/Open
04-acknowledgement.pdf394.1 kBAdobe PDFView/Open
05-chapter-1.pdf5.1 MBAdobe PDFView/Open
06-chapter-2.pdf5.65 MBAdobe PDFView/Open
07-chapter-3.pdf4.04 MBAdobe PDFView/Open
08-chapter-4.pdf2.98 MBAdobe PDFView/Open
09-chapter-5.pdf2.45 MBAdobe PDFView/Open
10-chapter-6.pdf2.92 MBAdobe PDFView/Open
11-chapter-7.pdf484.69 kBAdobe PDFView/Open
12-references.pdf3.06 MBAdobe PDFView/Open
13-publications.pdf242.82 kBAdobe PDFView/Open


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