بهبود كارايي مدل بازنمايي تنك براي بازگرداني و بازسازي تصوير

چكيده انگليسي :

Performance Improvement of Sparse Representation Model for Image Restoration and Reconstruction Mansour Nejati mansour nejati@ec iut ac ir Date of Submission 2016 09 06 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 IranSupervisor Shadrokh Samavi samavi96@cc iut ac irAdvisor Nader Karimi nader karimi@cc iut ac irDepartment Graduate Program Coordinator Mohammad Reza Taban Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan IranAbstractData models are a cornerstone of contemporary methodology in the field of signal and image processingwhich have evolved over the years In that respect the past decade has been certainly the era of sparse andredundant representations a popular and highly effective data model The main objective of this dissertationis the performance improvement of this model in image restoration and reconstruction To this end we firstconsider the performance improvement of sparse coding for noisy images for which low rank and nonlocalsparse representation models are proposed These models leads to superior image denoising performancecompared to the state of the art methods Second we concentrate on the dictionary as the core component ofsparse representation model To improve the performance of learned dictionaries for sparse representation acoherence regularized dictionary learning model is presented and two novel dictionary optimizationalgorithms are proposed Furthermore we propose a boosted dictionary learning approach to train adictionary ensemble which results in more efficient sparse representations Lastly a joint sparse model ispresented to train multiple dictionaries from different datasets taking into account their relationships Key WordsSparse representation dictionary learning sparse coding image restoration optimization IntroductionMuch of the progress made in image processing in the past decades can be attributed tobetter modeling of image content and a wise deployment of these models in relevantapplications Indeed a careful study of the image processing literature reveals that there isan evolution of such models and their use in practice In recent years a large amount ofmulti disciplinary research has been conducted on sparse and redundant representationmodeling with very successful results in a wide range of applications The fundamentalidea behind this model is a simple and compact representation of data with linearcombination of few basis elements or atoms from a set called a dictionary The sparsityas the driving force of this model has led to highly effective algorithms in many inverseproblems in image processing such as denoising inpainting super resolution and more A fundamental question in practicing the sparse representation model is the choice ofdictionary to be used In fact dictionary as the core ingredient of the model plays a criticalrole in a successful sparse representation modeling A recent approach to dictionary designis to learn overcomplete dictionaries from data that leads to specialized dictionary inrepresenting the signal in question Another important issue in processing of sparse signalsis sparse coding to obtain the sparse representation of a signal over a dictionary This

نجاتي، منصور

بهبود كارايي مدل بازنمايي تنك براي بازگرداني و بازسازي تصوير

يادگيري ديكشنري , كد گذاري تنك , بازگرداني تصوير , بهينه سازي