SEGMENTATION OF EXTREME ULTRAVIOLET (SOHO) SUN IMAGES BY MEANS OF WATERSHED AND REGION MERGING

摘要

The paper presents a new method of segmentation of Extreme UV images, which is based on the use of the watershed algorithm. The result of the watershed takes form of a partition of the whole image into a large number of segments. Some of these segments are then merged into larger regions representing meaningful objects, such as coronal holes. The proposed method of region merging finds the maximum of the average contrast between the current region and its boundary defined as a collection of watershed segments coming to touch with this region. The maximization of the average contrast gives the results which are in agreement with human intuition. Furthermore, this approach allows one to conduct the segmentation of the sun images almost independently of any tuning parameters. Tests of the method on sun images taken at various levels of sun's activity showed that the method can be used independently of the local brightness level and the extent of the coronal holes.