A set of morphological indicators is proposed to identify the characteristics of urban fabrics using spatial clustering. Local Indicators of Network-Constrained Clusters (LINCS) are preferred to classical Local Indicators of Spatial Association (LISA) in order to better integrate the point of view of pedestrians moving in the city. However, spatial analysis of the morphological indicators need careful consideration of their statistical asymmetries and of heteroscedasticity. Morphological rates are for example calculated on extremely variable base populations. Classical empirical Bayesian correction used in epidemiology, with the spatial unit surface area as base population, seems unfit to the analysis of urban morphology, as spatial units depend from morphological phenomena. New empirical Bayesian corrections are thus proposed and tested on the case study of urban landscapes of the French Riviera. A new Bayesian correction which is a sublinear function of the base population proves better able to reduce rate heteroscedasticity for several morphological indicators.