7848

7313

كارشناسي ارشد

مخابرات

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1391

دوازده،71ص.: مصور،جدول،نمودار

ص.ع.به فارسي و انگليسي

محمدرضا احمدزاده

سعيد صدري

22/4/92

مهندسي برق و كامپيوتر

ID7313

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Texture classification with Virtual Hexagonal Architecture Mohammadbagher Nourian m nourian@cc iut ac ir Date of submission 2013 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Persian Abstract Texture can be defined as a function of spatial pixel intensity Research on texture is a very important task in computer vision and its applications Texture recognition in human vision is easily can be defined but in machine vision and image processing has its special complexities All the system in this field is based on square architecture so the algorithms are faced with some problems such as intricacy huge amount of calculation and inadequate 40 years ago was brought a novel hexagonal architecture method in image processing Due to the advantages of this structure compared to square architecture soon attracted the attention of many researchers Some important advantages are equaled distances between one pixel and its adjacent pixels isotropy and good quality in edge and corner detection In this thesis a state of the art method for texture extraction in hexagonal structure is presented A completed local binary pattern CLBP in hexagonal structure is applied for information extraction In this method the texture information of each pixel express with three parameters first the magnitude local difference between central pixel and its neighbors second the sign local difference between central pixel and its neighbors and third central gray level in each pattern A feature vector for each image is achieved by applying mentioned descriptor and a dictionary is created by classifying train feature vectors Then by utilizing this dictionary a learning model for a set of texture images with different classes is specified Extensive experiments on several datasets from renowned texture databases such as Outex and the Brodatz database show that classification based on hexagonal structure performs much better than square structure Keywords Texture extraction Recognition Machine vision Hexagonal structure Isotropy A completed local binary pattern in hexagonal structure CLBPH Classification PDF created with pdfFactory trial version www pdffactory com

محمدرضا احمدزاده

سعيد صدري