Remote assessment of flooded areas based on inter-spectral statistical features

The paper presents an efficient method for image segmentation based on inter-spectral information. Two classes of regions were considered: flooded and non-flooded. The images were captured by unmanned aerial vehicles (UAV), fix-winged type. Using features extracted from benchmark samples of each texture, flooded areas are automatically recognized from input test images. Our results show that the Haralick textural features extracted from the co-occurrence matrix computed between spectral components bring significant information that help in the process of texture classification, increasing the accuracy of image segmentation. Moreover, we propose a new algorithm for computing the co-occurrence matrix between spectral components that provides slightly faster execution times than a similar one existing in the literature; we implemented it, using Matlab, with satisfactory results.