Heat Kernel Smoothing of Scalar and Vector Image Data

Fan Zhang, Edwin R. Hancock

Abstract

This paper shows how the graph-spectral heat kernel can be used to smooth both gray-scale and color images. We represent images using weighted attributed graphs, and compute the associated Laplacian matrix. Diffusion across this weighted graph-structure with time is captured by the heat-equation, and the solution, i.e. the heat kernel, is found by exponentiating the Laplacian eigen-system with time. Image smoothing is effected by convolving the heat kernel with the image. The method has the effect of smoothing within regions, but does not blur region boundaries. Experiments and comparisons on standard images illustrate the effectiveness of the method.
BibTeX
@inproceedings{zhang2006heat,
  author = {Fan Zhang and Edwin R. Hancock},
  title = {{Heat Kernel Smoothing of Scalar and Vector Image Data}},
  booktitle = {ICIP},
  year = {2006},
  doi = {10.1109/icip.2006.312646},
}

Loading related content...

Figure