Image processing of correlated data by experimental design techniques
New classes of algorithms are developed for processing of two-dimensional image data imbedded in correlated noise. The algorithms are based on modifications of standard analysis of variance (ANOVA) techniques ensuring their proper operation in dependent noise. The approach taken in the development of procedures is deductive. First, the theory of modified ANOVA (MANOVA) techniques involving one- and two-way layouts are considered for noise models with autocorrelation matrix (ACM) formed by direct multiplication of rows and columns or tensored correlation matrices (TCM) stressing the special case of the first-order Markov process. Next, the techniques are generalized to include arbitrary, wide-sense stationary (WSS) processes. This permits dealing with diagonal masks which have ACM of a general form even for TCM. As further extension, the theory of Latin square (LS) masks is generalized to include dependent noise with TCM. This permits dealing with three different effects of m levels using only m{sup 2} observations rather than m{sup 3}. Since in many image-processing problems, replication of data is possible, the masking techniques are generalized to replicated data for which the replication is TCM dependent. For all procedures developed, algorithms are implemented which ensure real-time processing of images.
- Research Organization:
- Polytechnic Univ., Brooklyn, NY (USA)
- OSTI ID:
- 5662663
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
Annual Report: Carbon Capture Simulation Initiative (CCSI) (30 September 2013)
MO-PIS-Exhibit Hall-01: Imaging: CT Dose Optimization Technologies I