Feature Extraction in the form of Statistical Moments Extracted to Bins formed using Partitioned Equalized Histogram for CBIR
H. B. Kekre1, Kavita Sonawane2
1Dr. H. B. Kekre, Sr. Professor, Computer Engineering, NMIMS University/ MPSTME/ Vile Parle, Mumbai, India.
2Kavita Soanwane, Ph. D. Research Scholar, Computer Engineering, NMIMS University/ MPSTME/ Vile Parle, Mumabi, India.
Manuscript received on January 17, 2012. | Revised Manuscript received on February 05, 2012. | Manuscript published on February 29, 2012. | PP: 98-109 | Volume-1 Issue-3, February 2012. | Retrieval Number: C0202021312/2011©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: This Paper introduces a new method of feature extraction in terms of statistical moments Mean, Standard deviation, Skewness and Kurtosis into three different bin sizes 8, 27 and 64 based on partitioned equalized histogram of the R, G, and B planes for content based image retrieval. Various feature vector databases are prepared and tested in this work to test response of the system through all small possibilities used in the feature extraction process based on invariant features. The system is designed to work with 2000 BMP images which include 20 different classes where each class has 100 images. Comparison process is core part of all CBIR systems; this system makes use of two similarity measures named Euclidean and Absolute distance for this purpose. System performance is evaluated using PRCP in addition to that LIRS, LSRR along with the newly introduced parameter ‘LONEGST String’ in the response of the given query for all the algorithms. Further the results obtained are refined and combined using the three criteria.
Keywords: Absolute distance, Equalized Histogram, Euclidean distance, LISR, LSRR, ‘Longest String’, PRCP.