Abstract
A new segmentation algorithm for the forest fire image based on Fractional Brownian motion and Contour let transform is proposed. Firstly, the algorithm decomposes the image into several sub-bands of multi-scale, location and multi-direction by Contour let transform. Fractional Brownian motion model is performed for low frequency sub-band, and then estimate the Hurst coefficient of each frequency coefficient in the sub-band. Hurst coefficient used as the characteristics of flame parameters for image segmentation. Experimental results show that the algorithm can segment flame effectively, and improve the efficiency of segmentation for forest fire image.