The data structure used to encode an image partition is of critical importance for most of region-based segmentation algorithms. Usual data structures are often convenient to extract only few parameters from the partition while inducing complex processing to compute others. Moreverer, the split and merge operations allowed by such data structure are often restricted. A new model~itebraquelaire-96 allows segmentation algorithms to extract a wide range of parameters from a partition. In this paper we describe the two basic primitives used by segmentation algorithms to modify a partition: the segment insertion and segment suppression.