Open AccessBinary Wavelet Transform Based Histogram Feature for Content Based Image Retrieval
Author(s) -
Megha Agarwal,
R. P. Maheshwari
Publication year - 2011
Publication title -
international journal of electronic signal and systems
Language(s) - English
Resource type - Journals
ISSN - 2231-5969
DOI - 10.47893/ijess.2011.1014
Subject(s) - artificial intelligence , pattern recognition (psychology) , wavelet transform , discrete wavelet transform , content based image retrieval , correlogram , computer science , color histogram , computer vision , rgb color model , wavelet , mathematics , image retrieval , color image , image processing , image (mathematics)
In this paper a new visual feature, binary wavelet transform based histogram (BWTH) is proposed for content based image retrieval. BWTH is facilitated with the color as well as texture properties. BWTH exhibits the advantages of binary wavelet transform and histogram. The performance of CBIR system with proposed feature is observed on Corel 1000 (DB1) and Corel 2450 (DB2) natural image database in color as well as gray space. The results analysis of DB1 database illustrates the better average precision and average recall of proposed method in RGB space (73.82%, 44.29%) compared to color histogram (70.85%, 42.16%), auto correlogram (66.15%, 39.52%) and discrete wavelet transform (60.83%, 38.25%). In case of gray space also performance of proposed method (66.69%, 40.77%) is better compared to auto correlogram (57.20%, 35.31%), discrete wavelet transform (52.70%, 32.98%) and wavelet correlogram (64.3%, 38.0%). It is verified that in case of DB2 database also average precision, average recall and average retrieval rate of proposed method are significantly better.