In order to use structural techniques from graph-based pattern recognition, a first necessary step consists in extracting a graph in an automatic way from an image. We propose to extract plane graphs, because of algorithmic properties these graphs have for isomorphism elated problems. We also consider the problem of extracting semantically well-founded graphs as a compression issue: we get simple graphs from which can be rebuilt images similar to the initial image. The technique we introduce consists in segmenting the original image, extracting interest pixels on the segmented image, then converting these pixels into pointels, which in turn can be related by region-based triangulation. We show the feasibility and interest of this approach in a series of experiments.