Rough Fuzzy Measures in Image Segmentation and Analysis - Rough Fuzzy Image Analysis: Foundations and Methodologies - page 227

and DB correlations decrease with the increase of the data dimensionality. From the exper-
imental data the the assumption of the best correlation and solutions have been observed
for CCD and FCD RECA populations for β-index and FFT and FFD RECA populations
for Dunn and DB indices. The further research in the rough entropy area should help to
better confirm and understand presented tendencies.
Proposed by the authors Rough Entropy Clustering Algorithm with different rough en-
tropy measures makes use of boundary region information during clustering process, and
at the same time comprehend better internal data structure. RECA algorithmic scheme
incorporates fuzziness and uncertainty into clustering procedure making possible a deeper
look into internal data structure. In this context, the careful insight into similarities and dif-
ferences between standard data analysis techniques and the new proposed approach seems
to be crucial in order to properly explore and apply this fuzzy-rough algorithms. In order
to answer the question of advantages of fuzzy rough data analysis in the area of image
segmentation, the paper concentrates on discovering present relations between established
data clustering validation measures (indices) and rough entropy measures. In this way, the
area for application of the proposed solution becomes clearly more evident and helps better
understand the nature of fuzzy-rough data analysis.
Acknowledgments
The research is supported by the grants N N516 0692 35 and N N516 3774 36 from the
Ministry of Science and Higher Education of the Republic of Poland. Computational exper-
iments were performed on a cluster built by the Department of Computer Science, Bialystok
University of Technology.