The city is a complex “object” whose structure can be studied at several levels of scale. In this article, we propose to work on the scale of the street, the one in which the pedestrian is immersed, and more precisely on the scale of its articulations, the street intersections. These are indeed structuring places, variously walkable, potentially difficult to cross, which a methodical description can facilitate the use. To this end, we operate, in each intersection, a matching between the shape of the open space as captured by the pedestrian in immersion (this visual pattern is more commonly called the isovist) and a corpus of geospatial patterns. This matching exploits a wavelet compression technique from signal processing which also has the advantage of evaluating the orientation of the pattern. The different urban fabrics presented during the comparative analysis highlight the versatility of the method but also its scalability.