The binary nature of document degradation decides the suitability of morphological methods for restoration. Although the computational burden in morphological filter design can be mitigated by imposing constraints on the filter and employing the morphological filter MAE theorem in an efficient search strategy, the design constraints on the filter limit the performance of single-pass filter. It has been shown that iterative morphological filters can outperform single-pass filters. The investigation of iterative morphological filter design for image restoration is the main contribution of the present thesis. The study of iterative morphological filter design provides the understanding in depth of how filters achieve a better restoration in an iterative way. Various image-noise processes have been used to examine the effect of iteration on window constraint. Through iteration we have increased the class of filters from which an increasing estimator may be designed, so that the window constraint can be compensated by employing iterative morphological filter. Practically, we arrive at the conclusion that smaller size observation windows can achieve very similar restoration result in a MAE sense as large size windows by employing iterative design. It provides us a better practical design of increasing operators for document restoration compared to the single-pass filter using large size window. Theoretically, we arrive at the conclusion that it is not important if two operators are quite different in logical structure, and they can achieve very similar restoration effect as long as they are statistically similar.
Library of Congress Subject Headings
Image processing--Mathematics; Digital filters (Mathematics); Morphisms (Mathematics)
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Zhang, Yeqing, "Iterative morphological filters and application in document restoration" (1995). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus