Klasifikace obrázků založená na rysech

Abstract

The aim of this master‘s thesis is to create appropriate image searching system based on classification according to specific features.

The first part focuses to existing methods that are useful for total image analysis that helps us to assign features for next user‘s searching.

The main part of a thesis is my own implementation. There are described all processes theoretically and practically. Part of it is annotation data set creation, color shades development for searched colors and image division into specific categories for their sorting. We get familiar with image segmentation connected with assigning of it‘s parts to trivial objects. Last but not least, we evaluate and save final features into database.

In conclusion is described an experiment that measures own implementation quality on annotation data set where we can consider some method changes or improvements.