engleski

Color based region classification in Mediterranean landscape images

Color is one of the most useful features when it comes to segmentation and classification of regions on images. As there are various color spaces, the question is which one is the most suitable for these tasks. Of course the answer depends of the region we want to classify and the image types. It is not the same if we have indoor or outdoor images. The main focus of this presentation is color-based segmentation and classification of regions in single images of Mediterranean landscape captured mostly by wildfire video based surveillance and monitoring systems. Therefore classes have been specially chosen in order to improve algorithms for wildfire smoke detection. The main goal of this presentation is to discover which color spaces, out of the nine different ones that we tested, are most suitable for the color-based classification of 11 different classes that can usually be found in natural Mediterranean landscape images. Additionally, we will also present a FESB MLID (FESB is a Croatian acronym for the Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture and MLID means Mediterranean Landscape Image Dataset) that we have collected during our research and make public for all researchers.

region classification; image analysis

nije evidentirano