Authors:
Julien Chauveau
;
David Rousseau
;
Paul Richard
and
François Chapeau-Blondeau
Affiliation:
Université d’Angers, France
Keyword(s):
Color image, Color histogram, Fractal, Self-similarity, Capacity dimension, Correlation dimension, Pair correlation integral, Feature extraction and analysis, Image modeling, Virtual reality, Vision.
Related
Ontology
Subjects/Areas/Topics:
Computer Vision, Visualization and Computer Graphics
;
Fractal and Chaos Theory in Image Analysis
;
Image and Video Analysis
Abstract:
The colorimetric organization of RGB color images is analyzed through the computation of algorithms which can characterize fractal organizations in the support and population of their three-dimensional color histogram. These algorithms have shown that complex organizations across scales exist in the colorimetric domain for natural images with often non-integer fractal dimension over a certain range of scale. In this paper, we apply this method of colorimetric characterization to synthetic images produced by rendering techniques of increasing sophistication. We show that the fractal or scale invariant signatures are more pronounced when the realism of the synthetic images increases. Such results could have interesting applications to improve the colorimetric realism of synthetic images. This also may contribute to progress in classification and vision, in using fractal colorimetric properties to differentiate natural and synthetic images.