Gaussian Blurring Mean Shift segmentace obrazu pomocí technologie NVIDIA CUDA

Abstract

Image segmentation is one of the basic pillars of computer vision and image processing. Mean-Shift clustering method is the computationally challenging one, providing space for parallel algorithm design. The goal of this bachelor thesis is to become familiar with the state of the art segmentation methods and NVIDIA CUDA technology. The practical part of this thesis is focused on design and efective implemetation of Gaussian Blurring Mean-Shift algorithm for processing both the CPU and GPU using CUDA technology. Thesis contains a comparison of the various implementations both in terms of achieved performance and difficulty of their development.