HomeChair: Karen O. Egiazarian, Tampere University of Technology, Finland
HomeEdward R. Dougherty, Texas A&M University (U.S.A.)
Granulometric transforms are used for image (texture) classification and characterization of optimal morphological (granulometric) bandpass filters. The present paper discusses parallels between these two applications of granulometric transforms for random sets and two corresponding applications of Fourier transforms, classification of random arcs by Fourier descriptors and characterization of the Wiener filter via power spectral densities.
ns970411.pdf (Scanned) ns970411.pdf (From Postscript)
TOPL. Yaroslavsky, Tel Aviv University (Israel)
A family of local adaptive filters for image restoration and enhancement is described. The filters work in a moving window in the domain of an orthogonal transform and, in each position of the window, produce an estimation of the central pixel of the window by nonlinear transformations of the window spectral coefficients. The filter synthesis on the base of local mean squared error restoration criterion is outlined, the use of Discrete Fourier and Discrete Cosine Transforms as orthogonal transforms is justified and recursive algorithm of multicomponent local DCT spectral analysis is described. Performance of filtering is illustrated by examples of edge preserving noise suppression and blind restoration of color images
ns970412.pdf (Scanned) ns970412.pdf (From Postscript) TOP
Robert Nowak, Michigan State University (U.S.A.)
Richard Baraniuk, Rice University (U.S.A.)
In this paper we describe two new structures for nonlinear signal processing. The new structures simplify the analysis, design, and implementation of nonlinear filters and can be applied to obtain more reliable estimates of higher-order statistics. Both structures are based on a two-step decomposition consisting of a linear orthogonal signal expansion followed by scalar polynomial transformations of the resulting signal coefficients. While most existing approaches to nonlinear signal processing characterize the nonlinearity in the time domain or frequency domain; in our framework any orthogonal signal expansion can be employed. In fact, there are good reasons for characterizing nonlinearity using more general signal representations like the wavelet transform. Wavelet expansions often provide very concise signal representation and thereby can simplify subsequent nonlinear analysis and processing. Moreover, we show that the wavelet domain offers significant theoretical advantages over classical time or frequency domain approaches to nonlinear signal analysis and processing.
ns970413.pdf (Scanned) ns970413.pdf (From Postscript) TOP
Jaakko T. Astola, Tampere University of Technology (Finland)
Karen O. Egiazarian, Tampere University of Technology (Finland)
Rusen Oktem, Tampere University of Technology (Finland)
Sos Agaian, University of Texas at San Antonio (U.S.A.)
In this paper we introduce parametric binary Rademacher functions of two types. Based on them using different kinds of arithmetical and logical operations we generate a set of binary polynomial transforms (BPT). Some applications of BPT in nonlinear filtering and in compression of binary images are presented.
ns970414.pdf (Scanned) ns970414.pdf (From Postscript) TOP
Karen O. Egiazarian, Tampere University of Technology (Finland)
Jaakko T. Astola, Tampere University of Technology (Finland)
Samvel M. Atourian, Tampere University of Technology (Finland)
David Z. Gevorkian, Tampere University of Technology (Finland)
A combination filter structure involving wavelet transform based denoising methods and K-nearest neighbor (K-NN) type operations is proposed and studied. Performance analysis of this filter shows its high efficiency in suppressing mixed white Gaussian and impulsive noises. At the same time the proposed filter possess moderate computational complexity.
ns970415.pdf (Scanned) ns970415.pdf (From Postscript) TOP