This work introduces two variants of unsupervised color segmentation methods. The underlying idea is to segment the input image several times, each time focussing on a different salient part of the image and to subsequently merge all obtained result