Bootstrap coverage plots for image segmentation

Two methods for assessing the variability of image segmentation are proposed: coverage plots and percentile intervals. Both methods are based on bootstrap techniques and provide graphical information about the performance and accuracy of segmentation methods across the entire image. The methods are applied to median based image segmentation which is characterised by the filter size and a data-driven thresholding choice. Computer simulations of simulated medical X-ray images demonstrate the superiority of coverage plots in assessing the performance of segmentation methods.