Segmentation of Photospheric Solar Images by Using c-Means, k-Means, and FCM Algorithms

In this study, we use three kinds of clustering methods based on c-means, k-means, and fuzzy c-means (FCM) algorithms to segment solar ultra-violet (UV) images. The methods are applied on a sequence of quiet-Sun photospheric observations at 525 nm images taken by Sunrise on 9 June 2009. The comparison between these three algorithms represents a little bit differences in extraction of physical parameters (filling factors, brightness fluctuations, size distribution, etc.) from images. On the basis of FCM algorithm, the mean value of granule sizes is found to be about 1.8 arcsec2 (0.85 Mm2). Granules with sizes smaller than 2.8 arcsec2 cover a wide range of brightness, while larger granules approaches a particular value. Granules may have lifetimes less than 10 minutes in this part of the Sun. Investigation of local fractal dimension of photospheric images shows that granulation pattern are approximately scale free in some resolutions.