Detekce významných oblastí v obrazech

Abstract

This bachelor's thesis studies the detection of salient regions in images. When observing the environment, one is able to distinguish areas of interest from an unimportant background. Automatic detection of such areas could help in many image analysis applications, such as speeding up the detection and recognition of objects in images. The aim of this work is to analyze and compare the behavior of three known methods for the detection of salient regions and to verify their success on a selected set of images. In the first part of this work we present an overview of existing methods for salient region detection in images, then we describe and analyze three of these methods. Finally, we test these methods on a set of images from the public MSRA10K dataset and compare the results.