This paper deals with the dynamic neuronal approach for segmentation of textured seafloors from sidescan sonar imagery. For classical approaches of sonar images segmentation, the result of the classification is a set of sediment clusters representing the different kinds of seabed. However, those classical approaches give satisfying results only when a comprehensive training set is available. If the training set lacks a particular kind of seabed, it will be unknown for the classifier and the classification will be reduced to the closest known sediment cluster. As it is not always feasible to know the entire seabed types before the training phase, a dynamic algorithm solution capable of incremental learning has been developed. The Dynamic Self-organizing maps (DSOM) algorithm used in this work is an extension version of classical SOFM (Self-Organizing Feature Map) algorithm developed by Kohonen combined with Adaptive resonance Theory (ART). It is based on growing neuronal map size during the learning processes. Therefore, the size of the map is small in the beginning but increase dynamically using control vigilance threshold. To assess the consistency of the proposed approach, the DSOFM algorithm is tested on simulated data clusters and on real sonar data.