Academic Journals Database
Disseminating quality controlled scientific knowledge

Image Retrieval using Fractional Energy of Row Mean of Column Transformed Image with Six Orthogonal Image Transforms.

ADD TO MY LIST
 
Author(s): H. B. Kekre | Sudeep D. Thepade | Archana A. Athawale | Paulami Shah

Journal: International Journal of Soft Computing & Engineering
ISSN 2231-2307

Volume: 1;
Issue: 4;
Start page: 168;
Date: 2011;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: CBIR | Cosine Transform | Walsh Transform | Haar Transform | Sine Transform | Slant Transform | Hartley Transform | Fractional Coefficients | Row Mean.

ABSTRACT
The thirst of better and faster retrieval techniqueshas always fuelled to the research in content based imageretrieval (CBIR). The paper presents innovative content basedimage retrieval (CBIR) techniques based on feature vectors asfractional coefficients of row mean of column transformedimages using Discrete Cosine, Walsh, Haar, Slant, Discrete Sine,and Hartley transforms. Here the advantage of energycompaction of transforms in low frequency coefficients intransform domain is taken to greatly reduce the feature vectorsize per image by taking fractional coefficients of row mean ofcolumn transformed image. The feature vectors are extracted insix different ways from the transformed image, with the firstbeing considering all the coefficients of row mean of columntransformed image and then six reduced coefficients sets (as50%, 25%, 12.5%, 6.25%, 3.125%, 1.5625% of complete rowmean of column transformed image) are considered as featurevectors. The six transforms are applied on the colourcomponents of images to extract row mean of columntransformed RGB feature sets respectively. Instead of using allcoefficients of transformed images as feature vector for imageretrieval, these six reduced coefficients sets for RGB planes areused, resulting into better performance and lower computations.The proposed CBIR techniques are implemented on a databasehaving 1000 images spread across 10 categories. For eachproposed CBIR technique 40 queries (4 per category) are firedon the database and net average precision and recall arecomputed for all feature sets per image transform. The resultshave shown performance improvement (higher precision andrecall values) with fractional coefficients compared to completetransform of image at reduced computations resulting in fasterretrieval. Finally Discrete Cosine Transform (DCT) surpassesall other discussed transforms in performance with highestprecision and recall values for 50% of fractional coefficients.
Why do you need a reservation system?      Save time & money - Smart Internet Solutions