Level Set Segmentation of Oil Spills from Earth Observatory Images Via Spatial KFCM Clustering

Abstract

In this paper, we present a novel technique called spatial kernel fuzzy clustering with adaptive level set approach for Oil spill image segmentation. The proposed method is diversified into two stages; in the first stage the input is pre-processing by Spatial Kernel Fuzzy C-Means clustering (KFCM) to improve the clustering efficiency and less sensitive to noise. In the second stage, it necessary to use the level set method to refine the previous stage segmentation results. The performance of the level set segmentation is subjected to proper initialization and optimal formation of directing parameters. The controlling parameters of level set evolution are also projected after the results of kernel fuzzy clustering. The proposed method, spatial kernel fuzzy adaptive level set algorithm is enhanced the local minima problem. Such developments enable level set handling and more strong segmentation. The results confirm its effectiveness for oil spill images over the conventional CV model i.e number of iterations, Computational time and PSNR