Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Comp Stat

Kolmogorov–Zurbenko filters

Full article on Wiley Online Library:   HTML | PDF

We reviewed the properties of the Kolmogorov–Zurbenko (KZ) filter and its extensions with applications in high resolution signal and image processing. The KZ filter is defined as an iteration of a moving average (MA) filter. The impulse response function of the KZ filter is a convolution of the rectangular window being used in a MA filter. Zero derivatives at the edges of the impulse response function make it a sharply declining function, providing high frequency resolution. The KZ Fourier transform (KZFT) is derived from the KZ filter by applying it to Fourier transform. Extensions of the KZ filter and the KZFT are demonstrated with examples. Copyright © 2010 John Wiley & Sons, Inc.

Figure 1.

Comparisons of the MSE as a function of the underlying smoothness of the spectrum for different spectral windows: o(α) is from the optimal spectral window; k(α) is from the KZ window; p(α) is from Parzen window; u(α) is from rectangular window; t(α) is from Tukey–Hamming window; b(α) is from Bartlett window. For the exact formula of MSE, please see Ref 1.

[ Normal View 28K | Magnified View 42K ]
Figure 2.

Absolute value of the impulse response function of KZ(21,5) and KZ(21,1).

[ Normal View 3K | Magnified View 7K ]
Figure 3.

Example of spatial KZ filter: (a) raw data; (b) one iteration of spatial 5 × 5 moving block average; and (c) two iterations of spatial 5 × 5 moving block filter.

[ Normal View 33K | Magnified View 115K ]
Figure 4.

Example of KZA filter to detect breaks in image data: (a) raw data and (b) filtered data.

[ Normal View 39K | Magnified View 61K ]
Figure 5.

10 log10 of the energy transfer function of KZFT of m = 201, k = 1 (black) and k = 5 (red) at frequency 0.025.

[ Normal View 32K | Magnified View 57K ]
Figure 6.

Example using KZFT to reconstruct signal.

[ Normal View 7K | Magnified View 19K ]
Figure 7.

Tidal wave was reconstructed using the KZFT and shifted forward 5 hours. Wind speed was smoothed using KZ(3 hours, 2) filter. Data are from August 23, 2005 19:00 to August 28, 2005 23:00 local time, when hurricane Katrina was at doorsteps of New Orleans. Horizontal scale is using dates in August 2005 at the points of midnights local time.

[ Normal View 5K | Magnified View 12K ]
Figure 8.

Contour plot of time–frequency map of two close floating frequency chirp signals by KZFT.

[ Normal View 62K | Magnified View 131K ]
Figure 9.

Diagram of the parameters in the definition of KZP and KZTP.

[ Normal View 1K | Magnified View 2K ]
Figure 10.

KZP (solid line) and true spectral density function (dash line).

[ Normal View 3K | Magnified View 8K ]
Figure 11.

Plots of raw third-order periodogram with L = 500, m = 1000 and (a) k = 1; (b) k = 3. (c) and (d) are the adaptively smoothed version of (a) and (b), respectively.

[ Normal View 54K | Magnified View 92K ]

Related Articles

Imaging: An Interdisciplinary View

Browse by Topic

Applications of Computational Statistics > Signal and Image Processing and Coding
blog comments powered by Disqus

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts

Twitter: WIREsCompStat Follow us on Twitter