Stated preference choice experiments can benefit from the use of images to describe complex scenarios such as public spaces or urban infrastructure. However, images will be perceived subjectively by users who will probably understand them in a more qualitative than quantitative way. A method to quantify the relevance of qualitative attributes of public spaces such as beauty, safety or security is proposed. The method is based on the sequential estimation of discrete choice and latent variable models. Two case studies are described and the complexity of the construction of images is discussed. Results show that the proposed method allows the inclusion of qualitative features in choice models, but the use of images introduces a bias in the perception of the magnitude of design attributes that should be analysed carefully. (C) 2016 The Authors. Published by Elsevier B.V.