Diffusion-Based Similarity for Image Analysis
Measuring the distances is a key problem in many image-analysis algorithms. This is especially true for image segmentation. It provides a basis for the decision whether two image points belong to a single or to two different image segments. Many algorithms use the Euclidean distance, which may not be the right choice. The geodesic distance or the k shortest paths measure the distance along the surface that is defined by the image function. The diffusion distance seems to provide better properties since all the paths are taken into account. In this paper, we show that the diffusion distance has the properties that make it difficult to use in some image processing algorithms, mainly in image segmentation, which extends the recent observations of some other authors. We propose a new measure called normalised diffusion cosine similarity that overcomes some problems of diffusion distance. Lastly, we present the necessary theory and the experimental results.
KeywordsDistance measurement Euclidean distance Geodesic distance Diffusion distance Normalised diffusion similarity
This work was partially supported by the grant of SGS No. SP2015/141, VŠB - Technical University of Ostrava, Czech Republic.
- 2.Brand, M.: A random walks perspective on maximizing satisfaction and profit. In: Proceedings of the International Conference on Data Mining, pp. 12–19 (2005)Google Scholar
- 7.Huang, H., Yoo, S., Qin, H., Yu, D.: A robust clustering algorithm based on aggregated heat kernel mapping. In: International Conference on Data Mining, ICDM 2011, pp. 270–279 (2011)Google Scholar
- 10.Lipman, Y., Rustamov, R.M., Funkhouser, T.A.: Biharmonic distance. ACM Trans. Graph. 29, 1–11 (2010)Google Scholar
- 11.Martin, D., Fowlkes, C., Tal, D., Malik, J.: A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. In: Proceedings of the International Conference of Computer Vision, ICCV 2001, pp. 416–423 (2001)Google Scholar
- 15.Sharma, A., Horaud, R., Cech, J., Boyer, E.: Topologically-robust 3D shape matching based on diffusion geometry and seed growing. In: Proceedings of the Conference on Computer Vision and Pattern Recognition, pp. 2481–2488 (2011)Google Scholar
- 16.Sinop, A.K., Grady, L.: A seeded image segmentation framework unifying graph cuts and random walker which yields a new algorithm. In: Proceedings of the International Conference on Computer Vision, ICCV 2007, pp. 1–8 (2007)Google Scholar