Robust Methods for Signal Parameter Estimation and Image Segmentation

点击翻译按钮获取中文摘要

This project is concerned with statistical signal processing and image processing. In particular the PI is developing robust algorithms for signal parameter estimation in correlated noise and also for the segmentation of objects from background in noisy images. Robustness is important because otherwise optimal schemes can perform poorly in the presence of statistical outliers. The goal of robust signal processing is to maintain the near optimal performance in such a case. In the parameter estimation area, it is planned to concentrate on nonlinear estimation in nominally Gaussian noise, making use of previously developed results in linear estimation and estimation in independent and weakly dependent noise. In the image segmentation area, the concentration will be on image models that result in robust maximum a posteriori (MAP) segmentation algorithms. The three reviewers assembled in a panel rated this proposal as Excellent. The investigators career potential is thought to be outstanding. The area of research is quite relevant to the CSP program. A two-year award of $57924 is highly recommended. This award should be funded at $25088 in FY87 and $32836 in FY88.