Abstract
In this work, preliminary experiments were conducted on a novel method seeking physical parameters of 3D objects that might strongly contribute to their visual impressions through computational investigation of the impression ratings made on learning samples. Car body was selected as a class of 3D objects to be investigated, and a set of the 3D shape data for various car types was obtained by measuring plastic models of different car lines with a range-finder. Morphable 3D model of car bodies describing the variation of their appearance in terms of a smaller number of parameters was obtained by applying PCA to a set of high dimensional vectors representing the 3D shapes. A preliminary method was proposed to transform the visual impression of the 3D object by manipulating the parameters defined in the morphable 3D model. Results of psychological experiments conducted by applying the method of paired comparison indicated possible validity of the proposed method. A new scheme is also proposed to properly re-sample a novel object having even a peculiar shape so that such object could also be represented by the morphable 3D model.
