On a segmentation method to extract fire regions in fire scenes
Résumé
The aim of this study is to extract fire regions in images. At first a study is conducted on RGB, HSV, YUV and Lab color spaces in order to characterize the fire pixels distributions on each channel. A new method is then proposed based on k-means clustering in the b-channel of Lab and on distribution properties of fire pixels in the RGB color space. The performances of the proposed method are first qualitatively presented, then a quantitative evaluation is conducted using a supervised evaluation criterion based on expert annotations. A comparative study is performed with other existing algorithms assessing the good behavior of the presented method.