In this paper, we propose to compare various unsupervised algorithms of pixel classification, and this, in various color spaces, for the segmentation of color images. This article is composed of three parts described thereafter. In the first one, we clarify the data processing sequence of our color segmentation. In the second part, we detail the classification of pixels and color watershed methods. In the last part, we evaluate the results of the segmentation by the Liu's method on images of bronchial cytology. Finally, we conclude that our technique of segmentation is adapted to the extraction of the cytoplasm of the cells.