| Autor: |
K., Venkataravana Nayak, S. K., Sharathkumar, J. S., Arunalatha, K. R., Venugopal |
| Předmět: |
|
| Zdroj: |
IAENG International Journal of Computer Science; Jun2023, Vol. 50 Issue 2, p793-802, 10p |
| Abstrakt: |
Image retrieval plays an important role in the analysis of obtained decisive visual information. The presence of visual inconsistency in visual appearance decreases the retrieval accuracy and many of the present retrieval methods emphasize single-source retrieval with the assumption of queries and databases distributions being similar. The number of features obtained with the traditional approach in which some of them are redundant, correlated, and sometimes noisy, increases the model feature space complexity and decreases interpretability. From the study of previous work, it is evident that feature fusion with cross-domain retrieval has not been addressed thoroughly so far. Thus, to deal with these issues, this extracts the optimal combination of the multi-modal features and fuses for enhancing retrieval accuracy. The complementary features obtained are effective with the traditional approach for the improvement of representation and retrieval effectiveness. Thus, Image Retrieval using Single and Cross-Domain Feature Fusion (SCDFF) is proposed in this work. The multi-modal features are extracted with Texture, Color, Statistical, and Scale Invariant Feature Transform (SIFT) descriptors to perform the retrieval process. The feature vector is fused using an optimized weight value which is obtained from Glowworm Swarm Optimization (GSO) algorithm and the image similarity is computed with K-Nearest Neighbor. An empirical analysis is performed to evaluate the proposed model and from the results obtained, it is evident that this work outperforms existing approaches in terms of accuracy. The novelty of this work lies in the fact of Single-Domain Feature Fusion (SDFF) and Cross-Domain Feature Fusion (CDFF) with optimization for Image Retrieval. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
