Multiresolution Analysis with Pulses
Multiresolution analysis has recently received considerable attention in relation to wavelets. The word “multiresolution” is appropriate in so far as wavelets are local in some sense, and therefore have exponentially decaying impulse response. In image processing it is clear that edges and impulses yield undesirable synthetic features in partially reconstructed images from linear multiscale decompositions. Median decompositions are regarded as better in practice, but computational complexity and lack of theory are problems. An alternative, from mathematical morphology is possible, yielding results demonstrably similar to the median decomposition, but computationally simpler, and having a strong theory for deriving qualitative and quantitative properties.
KeywordsWavelet Decomposition Resolution Level Multiresolution Analysis Mathematical Morphology Impulsive Noise
Unable to display preview. Download preview PDF.
- A. Bijaoni, F. Murtagh, J.L. Starck, Image Processing and Data Analysis -The Multiscale Approach, Cambridge University Press, 1998.Google Scholar
- C.H. Rohwer, Fast Approximation with Locally Monotone Sequences. Proceedings 4th FAAT Conference, Maratea, (to appear).Google Scholar
- H.L. Royden, Real Analysis, Macmillan, N.J., 1969.Google Scholar
- J. Serra, Image Analysis and Mathematical Morphology, Acad. Press, London 1982.Google Scholar
- Zhou Xing-Wei, Yang De-Yun, Ding Run-Tao, Infinite length roots of median filters, Science in China, Ser. A, Vol. 35 (1992), 1496–1508.Google Scholar