Premium
Visualization of the Centre of Projection Geometrical Locus in a Single Image
Author(s) -
Stojaković V.,
Popov S.,
Tepavčević B.
Publication year - 2014
Publication title -
computer graphics forum
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.578
H-Index - 120
eISSN - 1467-8659
pISSN - 0167-7055
DOI - 10.1111/cgf.12253
Subject(s) - computer science , visualization , projection (relational algebra) , redundancy (engineering) , process (computing) , constraint (computer aided design) , set (abstract data type) , artificial intelligence , computer vision , cultural heritage , creative visualization , information retrieval , human–computer interaction , data mining , algorithm , mathematics , geometry , archaeology , history , programming language , operating system
Abstract Single view reconstruction (SVR) is an important approach for 3D shape recovery since many non‐existing buildings and scenes are captured in a single image. Historical photographs are often the most precise source for virtual reconstruction of a damaged cultural heritage. In semi‐automated techniques, that are mainly used under practical situations, the user is the one who recognizes and selects constraints to be used. Hence, the veridicality and the accuracy of the final model partially rely on man‐based decisions. We noticed that users, especially non‐expert users such as cultural heritage professionals, usually do not fully understand the SVR process, which is why they have trouble in decision making while modelling. That often fundamentally affects the quality of the final 3D models. Considering the importance of human performance in SVR approaches, in this paper we offer a solution that can be used to reduce the amount of user errors. Specifically, we address the problem of locating the centre of projection (CP). We introduce a tool set for 3D visualization of the CP's geometrical loci that provides the user with a clear idea of how the CP's location is determined. Thanks to this type of visualization, the user becomes aware of the following: (1) the constraint relevant for CP location, (2) the image suitable for SVR, (3) more constraints for CP location required, (4) which constraints should be used for the best match, (5) will additional constraints create a useful redundancy. In order to test our approach and the assumptions it relies on, we compared the amount of user made errors in the standard approaches with the one in which additional visualization is provided.