An efficient graph-theoretical approach for interactive mobile image and video segmentation


Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2013

Öğrenci: OZAN ŞENER

Danışman: ABDULLAH AYDIN ALATAN

Özet:

Over the past few years, processing of visual information by mobile devices getting more affordable due to the advances in mobile technologies. Efficient and accurate segmentation of objects from an image or video leads many interesting multimedia applications. In this study, we address interactive image and video segmentation on mobile devices. We first propose a novel interaction methodology leading better satisfaction based on subjective user evaluation. Due to small screens of the mobile devices, error tolerance is also a crucial factor. Hence, we also propose a novel user-stroke correction mechanism handling most of the interaction errors. Moreover, in order to satisfy the computational efficiency requirements of mobile devices, we propose a novel spatially and temporally dynamic graph-cut method. Conducted experiments suggest that the proposed efficiency improvements result in significant computation time decrease. As an extension to video sequences, a video segmentation system is proposed starting after an interaction on key-frames. As a novel approach, we redefine the video segmentation problem as propagation of Markov Random Field (MRF) energy obtained via interactive image segmentation tool on some key-frames along temporal domain. MRF propagation is performed by using a recently introduced bilateral filtering without using any global texture or color model. A novel technique is also developed to dynamically solve graph-cuts for varying, non-lattice graphs. In addition to the efficiency, segmentation quality is also tested both quantitatively and qualitatively; indeed, for many challenging examples, quite significant time efficiency is observed without loss of segmentation quality.