Enhancing The Capability of Color based CBIR by Voting of Color Histograms
Searching by image features in image database is very important on the Internet. This process is called content based image retrieval (CBIR). Among the features used, the color feature is very effective especially when being used in color images. In this paper, we propose a new algorithm that uses t...
Saved in:
Published in | 2017 27th International Conference on Computer Theory and Applications (ICCTA) pp. 4 - 10 |
---|---|
Main Authors | , , |
Format | Conference Proceeding |
Language | English |
Published |
IEEE
28.10.2017
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Searching by image features in image database is very important on the Internet. This process is called content based image retrieval (CBIR). Among the features used, the color feature is very effective especially when being used in color images. In this paper, we propose a new algorithm that uses the votes of RGB and HSV color histograms in order to enhance the capabilities of the color feature of the images. The experimental results show the superiority of the proposed voting algorithm compared to using any other color histogram individually. The proposed algorithm is tested on the well-known Corel database which characterized by its large image set. The F-score is enhanced using the voting algorithm in most cases of the tests. The average F-score of the proposed algorithm exceeds other algorithms by 6%. The proposed algorithm is rotation, scaling, and translation invariant (RST). Additionally, the effect of thresholding is evaluated on the searching process. The proposed algorithm is very efficient in small databases that outperforms other algorithms by 13% in the F-score. |
---|---|
DOI: | 10.1109/ICCTA43079.2017.9497165 |