Paper Details: Downloads: 1674
Serial Number: P1151512364
Title: Multistage VQ based Feature Vector for Effective CBIR
Authors: V. R. Khapli and A. S. Bhalchandra
Abstract: Vector quantization (VQ) technique provides an effective codebook as representation of an image. This codebook is also very much useful as image feature from the view point of content based image retrieval (CBIR). Thus, designing an efficient codebook is the most important task in VQ. There are already several algorithms published on how to generate a codebook. However these codebooks when used as feature vector for CBIR still lacks the effectiveness as far as retrieval accuracy is concerned. To improve the retrieval accuracy, more effective feature vector is needed. To extract effective features for improving CBIR performance, a two pass VQ (MVQ) based FV technique is proposed in this paper and tested on Wang’s generic color image datasets. Also a novel framework for Multistage VQ (MSVQ) based CBIR using distortion value in every code vector is proposed for extracting effective FV. The proposed systems proved to be promising for improving retrieval accuracy for generic database search.
Keywords: Vector Quantization, Content Based Image Retrieval, Multistage Vector Quantization
Journal/Conference: International Journal of Graphics, Vision and Image Processing
Volume: 15
Issue: 1
Submission Date: 3/20/2015 12:00:00 AM
Review Date: 3/31/2015 12:00:00 AM
Publishing Date: 4/5/2015 12:00:00 AM
Article Downloads: 1674
Download:

Facebook