Modified Spatial Kernel Fuzzy clustering by Median Membership filtering and Level Set Method for Automatic Segmentation of Oil Spill from Earth Observatory Images

Abstract

In this paper, we extant a unique hybrid method called Modified Spatial kernel fuzzy Level set Method (MS-KFLSM) for accurate and fully automatic segmentation of oil spill regions from Earth Observatory SAR images. The proposed algorithm is diversified into phases called pre-processing and post-processing on the image. In the first phase, preprocessing on input SAR images by incorporating the spatial information and median membership filtering into Kernel Fuzzy C-means (KFCM) clustering to improve clustering enactment, accelerate convergence and getting rid of noise but outliers and boundary leakages are present even after the pre-processing. Next phase, post-processing is required to apply on preprocessing results in directive to refine the segmentation results and overcome the difficulties in the previous stage using an adaptive level set method. In this way, the results of the grouping before processing will be affected by limit losses and outliers. To reduce these limitations in the first phase with the help of the adaptive level approach set in the second phase, it is necessary to obtain solid and superior segmentation results. The presentation of the level set segmentation is exposed towardsasuitable introduction and a perfect conformation of the control parameters. It can immediately evolve from preliminary segmentation through widespread clustering of the space nucleus. The parameters controlling the evolution of the set of levels are also foreseen by the consequences of the widespread grouping of the nucleus. The proposed technique is more advantageous through the problem of local minima; these improvements facilitate solid segmentation. The proposed algorithm is performed superior segmentation compared to conventional level set techniques in terms of the number of iterations, calculation time and PSNR.