A novel enhanced method of getting autocorrelogram in content-based image retrieval |
Posted on:2006-02-24 | Degree:M.C.Sc | Type:Thesis |
University:Dalhousie University (Canada) | Candidate:Qiao, Yu | Full Text:PDF |
GTID:2458390008472556 | Subject:Computer Science |
Abstract/Summary: | |
The color correlogram is an image feature extraction method of the color feature that incorporates spatial information with color. This method is much better than the color histogram method and other histogram refinement methods, such as Color Coherent Vector (CCV). The traditional method of getting an autocorrelogram feature does not use any pre-processing procedure before extracting the feature from a quantized image based on RGB color space. This paper proposes an enhanced method of getting an autocorrelogram feature that applies a convolution filter to pre-process the raw images first and then quantizes colors based on HSV color spaces. The new method of using convolution filers can also be applied to get the color histogram feature. This method is here proven to be a very effective approach to improve retrieval results. |
Keywords/Search Tags: | Method, Feature, Color histogram, Getting, Autocorrelogram |
|
Related items |