2016 · 20 citations · 16 references
ComputingComputer ScienceComputer VisionMachine VisionRemote SensingImage SegmentationInfrared Remote SensingObject DetectionObject RecognitionRemote Sensing Image AnalysisSemantic SegmentationRegion Of Interest DetectionSaliency DetectionSaliency-based SearchSaliency-based SegmentationSalient Region DetectionThresholdingVisual SaliencyVisual Saliency Detection
In this paper we have worked on the problem of automatic ship detection in IR images. Segmentation of IR ship images is always a challenging task because of the intensity inhomogeneity, sea clutters and noise. In our proposed approach, we have shown that efficiency and accuracy of IR ship detection algorithms can be enhanced by only searching around the salient parts of the input image. In order to identify most salient regions, at first we computed the saliency map of the input image using the Graph-Based Visual Saliency (GBVS) algorithm. Next a multilevel thresholding of the saliency map is performed to get the ranked salient regions of the input image. By using the ship size as prior information, top-k regions are further processed to get the fine segmentation of the target. For this purpose we have used spatial constraint based fuzzy c-mean (FCM) segmentation algorithm along with a strategy to choose the cluster selection threshold. Experiments are performed on a data set of 18 diverse and challenging IR ship images, collected from different sources. Results show that our proposed framework is very effective and perform better compare to the methods which directly search the target in entire image.
16
Tony F. Chan, Luminita A. Vese · IEEE Transactions on Image Processing · 2001 · 10.3K citations
Josef Kittler, J. Illingworth · Pattern Recognition · 1986 · 2.2K citations