Mode Seeking with an Adaptive Distance Measure
The mean shift algorithm is a widely used non-parametric clustering algorithm. It has been extended to cluster a mixture of linear subspaces for solving problems in computer vision such as multi-body motion segmentation, etc. Existing methods only work with a set of subspaces, which are computed from samples of observations. However, noises from observations can distort these subspace estimates and influence clustering accuracy. We propose to use both subspaces and observations to improve performance. Furthermore, while these mean shift methods use fixed metrics for computing distances, we prefer an adaptive distance measure. The insight is, we can use temporary modes in a mode seeking process to improve this measure and obtain better performance. In this paper, an adaptive mode seeking algorithm is proposed for clustering linear subspaces. By experiments, the proposed algorithm compares favorably to the state-of-the-art algorithm in terms of clustering accuracy.
KeywordsMean Shift Algorithm Metric Learning
Unable to display preview. Download preview PDF.
- 4.Çetingül, H., Vidal, R.: Intrinsic mean shift for clustering on stiefel and grassmann manifolds. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1896–1902 (2009)Google Scholar
- 6.Vidal, R.: A tutorial on subspace clustering. IEEE Signal Processing Magazine (to appear)Google Scholar
- 7.Ho, J., Yang, M.H., Lim, J., Leem, K.C., Kriegman, D.: Clustering appearances of objects under varying illumination conditions. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 11–18 (2003)Google Scholar
- 8.Turaga, P., Veeraraghavan, A., Chellappa, R.: Statistical analysis on stiefel and grassmann manifolds with applications in computer vision. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1–8 (2008)Google Scholar
- 10.Sheikh, Y., Khan, E., Kanade, T.: Mode-seeking by medoidshifts. In: IEEE International Conference on Computer Vision, pp. 1–8 (2007)Google Scholar
- 11.Absil, P.A., Mahony, R., Sepulchre, R.: Optimization algorithms on matrix manifolds. Princeton University Press (2007)Google Scholar
- 13.Leibe, B., Schiele, B.: Analyzing appearance and contour based methods for object categorization. In: IEEE Conference on Computer Vision and Pattern Recognition, vol. 2, pp. 409–415 (2003)Google Scholar
- 15.Georghiades, A., Belhumeur, P., Kriegman, D.: From few to many: Generative models for recognition under variable pose and illumination. In: IEEE Computer Society Conference on Automatic Face and Gesture Recognition, pp. 277–284 (2000)Google Scholar