Cross-section estimation and image segmentation by simulated annealing

To cope with the problem of data inhomogeneity one requires some form of automatic parameter estimation and, if the data parameterisation allows, some form of segmentation. Ideally this parameter or cross-section estimation algorithm should be capable of adapting to the local statistics in such a way that it may cope with potentially different parameterisations for different image resolutions. An algorithm based on simulated annealing which is capable of providing accurate estimations of either radar cross-section, at high resolution, or other texture parameters at lower resolution is presented. Example segmentations produced by these smoothing operations are also presented. These results are presented visually