Computer graphics identification using genetic algorithm

Document Type

Conference Proceeding

Publication Date

1-1-2008

Abstract

This paper proposes the use of genetic algorithm to select an optimal feature set for distinguishing computer graphics from digital photographic images. Our previously developed approach has derived a 234-D feature vector from each test image in HSV color space. The statistical moments of characteristic functions of the image and its wavelet subbands were selected as the distinguishing image features. Since it is possible that only certain image features contain significant information with respect to the classification, the image features with insignificant contributions to classification may be eliminated to reduce the dimensionality of the feature vectors while maximizing the classification performance. Famous for its efficiency in searching the optimal solution in a very large space, the genetic algorithm is applied to find a reduced feature set which consists of only 100-D features per image in our investigation. The experimental results have demonstrated that the 100-D reduced feature set outperforms the 234-D full feature set. © 2008 IEEE.

Identifier

77957941324 (Scopus)

ISBN

[9781424421756]

Publication Title

Proceedings International Conference on Pattern Recognition

External Full Text Location

https://doi.org/10.1109/icpr.2008.4761552

ISSN

10514651

This document is currently not available here.

Share

COinS