Rank algorithms for picture processing

Structured nonlinear rank algorithms (RAs) for digital image processing are characterized and demonstrated. The fundamental principles and advantages of RAs are explained, and it is pointed out that several existing algorithms (involving adaptive amplitude transformations, local histogram equalization, and median or extremal filtration) are in effect members of this class. The effectiveness of RAs in picture smoothing, detail enhancement, and extraction of detail boundaries is illustrated in a series of sample images, and applications to automatic diagnostics of distortion and noise parameters, the statistical characterization of video signals, the measurement of texture features, and picture standardization and coding are considered.