13235

1130 دكتري

دكتري

مهندسي برق

اصفهان:‌ دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

۱۳۹۶

پانزده، [۱۳۰]ص.: مصور، جدول، نمودار

سعيد صدري، بهزاد نظري

ريحانه ريخته‌گران

محمدرضا احمدزاده، نادر كريمي

1396/11/24

كتابنامه

برق و كامپيوتر

مهندسي برق و كامپيوتر

ID1130 دكتري

Saliency Detection in Natural Images Based on Dirichlet Process Mixture Model Navid Rabbani Najafabadi n rabbani@ec iut ac ir January 9 2018 Department of Electrical Computer EngineeringIsfahan University of Technology 84156 83111 Isfahan IranSaeed Sadri sadri@cc iut ac irBehzad Nazari nazari@cc iut ac irReyhaneh Rikhtegaran Mohammad Reza Taban Department of Electrical Computer Engineering Isfahan University of Technology 84156 83111 Isfahan Iran Department of Mathematical Sciences Isfahan University of Technology 84156 83111 Isfahan IranAbstractThis dissertation introduces a new Bayesian framework for saliency detection In thisframework image saliency is computed as product of three saliencies location based feature based and center surround saliencies Each of these saliencies is estimated usingstatistical approaches The center surround saliency is estimated using Dirichlet processmixture model We evaluate our method using five different databases and it is shown thatit outperform state of the art methods KeywordsSaliency detection Bayesian framework Dirichlet process mixtureIntroductionIn general allocating a limited processing power of the brain to a specific part of visualdata is called visual attention Some features of a scene such as movement or colorcontrast attract visual attention unconsciously These characteristics are called imagesaliency Itti et al 1998 Regardless of the brain procedures for visual attention finding computational modelswhich can detect salient points in an image have found extensive applications in differentaspects of image processing and machine vision such as video surveillance objectdetection and recognition scene understanding advertisement image and videocompression and image retrieval These models are called saliency detection models Theiroutput is usually a saliency map comprised of an estimated saliency attached to each imagepixel Saliency roughly refers to the probability a pixel attracts visual attention There are two different mechanisms for human visual attention bottom up saliency andtop down saliency In bottom up mechanism or stimulus driven attention the low levelfeatures of an object are used to detect saliency In top down mechanism or goal driven

سعيد صدري، بهزاد نظري

ريحانه ريخته‌گران

محمدرضا احمدزاده، نادر كريمي