ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

SVD Based Features for Image Retrieval

Journal: International Journal of Computer Science and Artificial Intelligence (Vol.2, No. 2)

Publication Date:

Authors : ;

Page : 10-17

Keywords : CBIR; SVD; RGB; YCbCr; YUV; CXY; R’G’I; HSV; Precision; Recall; Euclidean Distance; Bray Curtis Distance;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

In this paper we present a new method for Content Based Image Retrieval (CBIR). Image signature computed by using the Singular Value Decomposition (SVD). Singular values used as a feature are obtained from SVD of full image and sub block of image with different color spaces. Seven color spaces are used for the proposed method. Singular values for the feature vectors are 8,16,32,64 and 200 for the full image and it is different for block based SVD. For block based SVD image we use 8x8, 16x16, 32x32, 64x64 and 128x128 sub blocks to calculate feature vector. So we can compare the result of different color with full and block based SVD. Similarity between the query image and database image measured here by using simple Euclidean distance (ED) and Bray Curtis distance (BCD). The average precision and average recall of each image category and overall average precision and overall average recall is considered for the performance measure. Proposed method tested on the database include 1200 images has 15 different classes to compare the performance.

Last modified: 2013-08-16 10:28:40