Image Segmentation using Various Local Structural Information

Abstract

Image segmentation is broadly used in computer vision applications. In literature, there are numerous techniques available for various applications. And yet, this field has many appli cations, where some challenging problems are unresolved. In this thesis, we tried to address some of these challenges by proposing a various model of object segmentation. In our work, we have considered two classes of problem, i.e., partial blur detection and multimodal object segmentation. newlineIn the first part, an interactive motion blur detection and segmentation of an object is carried out. Motion blur in an image occurs mainly due to relative motion of an object and cam era while acquiring a picture. Due to the motion of an object, each pixel in the resultant object is mixer of surrounding information. Henceforth, it is essential to restore such object for ap plications like, information retrieval, surveillance, medical imaging, bio-metric verification and so forth. newlineIn the second part, a selective segmentation techniques based on different characteristics of an object are investigated. Due to variation of the intensity in the object, invariably the variational segmentation model produces undesired segmentation results. To address this issue, an active contour model is proposed for selective segmentation of an object which has piecewise homogeneous region newline newline