Authors:
Nuru Jingili
1
;
Sigrid Ewert
1
and
Ian Sanders
2
Affiliations:
1
School of Computer Science and Applied Mathematics, University of the Witwatersrand, Johannesburg and South Africa
;
2
School of Computing, University of South Africa, Florida and South Africa
Keyword(s):
Perceptual Similarity, Similarity Measures, Picture Grammars, Random Context, Bag Context, Spatial Color Distribution Descriptor.
Related
Ontology
Subjects/Areas/Topics:
Case Studies
;
Conceptual Modeling
;
Formal Methods
;
Health Engineering and Technology Applications
;
Mathematical Simulation
;
Neural Rehabilitation
;
Neurotechnology, Electronics and Informatics
;
Simulation and Modeling
;
Simulation Tools and Platforms
Abstract:
This paper shows how similar pictures can be generated using random and bag context picture grammars. An online survey was conducted to determine the similarity of the pictures generated by the picture grammars. Respondents were asked to rank pictures in order of similarity to the query picture. They were also asked to rank galleries of pictures from those containing pictures that are most similar to those containing pictures that are least similar. Furthermore, respondents were required to tell us how they determined the similarity of the pictures contained in the galleries. We then compared perceptual similarity with a chosen similarity measure — spatial color distribution descriptor (SpCD) — to determine if they are consistent. The spatial color distribution descriptor has provided excellent results in determining the similarity of computer-generated pictures, and so was seen as a good similarity measure for this research. The results show that there is a good correlation between
the SpCD and perceptual similarity although in some instances humans do make different judgements.
(More)