Signal Detection and Block Truncation Coding in the Non-Gaussian Environment

摘要

A number of results were obtained pertaining to signal detection and data compression for image processing. These results led to improved performance over the previous approaches, with special attention given to methods which required less statistical knowledge and which were easier to implement. In particular, robustness and nonparametric techniques were employed to allow the exploitation of whatever knowledge was available, while retaining insensitivity to the remaining inexactness in knowledge. A number of results were developed pertaining to a new geometric measures of robustness which the researchers have orginated. This work has made possible the quantitative tradeoff between performance and robustness.