Title :
Nonlinear diffusion: A probabilistic view
Author :
Krim, H. ; Bao, Y.
Author_Institution :
Dept. of Electr. & Comput. Eng., North Carolina State Univ., Raleigh, NC, USA
Abstract :
A probabilistic view of diffusion is presented. A discrete symmetric random walk is shown to be equivalent to a heat equation evolution, and an extension to nonlinear evolutions including Perona-Malik equation is shown to be of utmost importance for analysis. Upon unraveling the limitations as well as the advantages of such an equation, we are able to propose a new approach which is demonstrated to outperform existing approaches, and to lift the long-standing problem of when to stop the evolution. Substantiating examples of image enhancement and segmentation are provided.
Keywords :
image enhancement; image segmentation; random processes; Perona-Malik equation; discrete symmetric random walk; image enhancement; image segmentation; nonlinear diffusion; nonlinear evolutions; probabilistic view; Filtering; Image analysis; Image enhancement; Image segmentation; Kernel; Nonlinear equations; Smoothing methods; Space heating; Stochastic resonance; Wavelet analysis;
Conference_Titel :
Image Processing, 1999. ICIP 99. Proceedings. 1999 International Conference on
Conference_Location :
Kobe
Print_ISBN :
0-7803-5467-2
DOI :
10.1109/ICIP.1999.822847
