Automatic segmentation and analysis of ancient mosaic images can help archeologists and experts build digital collections and automatically compare mosaics by means of image database indexing and content-based retrieval tools. However, ancient mosaics are characterized by low contrast colors, irregular tessella shape, orientation and positioning, making automatic segmentation difficult. In this work we propose a tessella-oriented strategy whose first step consists in isolating tessellas from its cemented network by computing the watershed transformation of a criterion image generated to exhibit the cement network as watershed crests. Then a simple k-means algorithm is used to classify tessellas and segment mosaic images with more accuracy than with a pixeloriented strategy. Additionally, we propose a method to automatically get the main directional guidelines of mosaics by estimating tessella orientation. This is done by minimizing a contextual energy computed from gray-level means of neighboring tessellas and orientation of their borders. Several examples of cartographies show the effectiveness of the method.