In this page i will share my academic studies with you. Hope it helps!
________________________________________________________________
Title: 3D Liver Vessel Segmentation
Abstract: Vessel segmentation is one of the challenging tasks in the field of image analysis. This paper gives a comprehensive study for liver vessel segmentation of 3D images. The main goal is to segment the liver vessels in a given 3D DICOM image of the abdomen. We proposed a cascaded liver segmentation structure such that we first extracted the liver region from the abdomen, then vessel enhancement is performed in order to refine the tubular structures in the extracted liver region. Then, we apply thresholding to detect initial contour which is later used for the initialization of the level set active contour. Lastly, we applied a level set active contour to segment the vessels inside the liver. In this study, we examined various techniques for each of the segmentation structure cascade such as vessel enhancement and level set active contour. Finally, we compared these techniques and select the most appropriate ones to build up our final segmentation structure.
Date added: 27/08/2011
Click here to open.
________________________________________________________________
Title: Space-Time Behavior-Based Action Recognition
Abstract: In this paper an action recognition approach is proposed which is based on space-time behavior-based similarity measure. This similarity measure can tell if two different video segments have the same similar motion field. It is calculated from intensity information. It allows recognizing actions that is performed by differently dressed people in different environments. Tests are done in KTH action dataset, which contains 6 different action classes with 25 different people performing actions on different environments with different clothes. Almost 80 % of recognition rate is achieved.
Date added: 27/08/2011
Click here to open.
________________________________________________________________
Title: Linear Isotropic & Nonlinear Anisotropic Diffusion Filtering on Images
Abstract: In image enhancement diffusion filtering methods are preferred, especially when the diffusion is nonlinear and the restoration process involves iterative evolutions of the degraded image and the diffusion is oriented with respect to the structural characteristics of the image. Basic idea is to enhance image by smoothing noisy pixels without losing the important data like edges. In this paper four different filtering methods are compared, the first three are L2-norm, L1-norm and surface area minimizing filters respectively, and the last one is coherence enhancing filter based on Weickert’s approach.
Date added: 27/08/2011
Click here to open.