Abstract | ||
---|---|---|
Content-based image retrieval (CBIR) became an energetic research field in engineering and medical field. In this work, local directional gradient based feature learning approach for image retrieval is proposed using artificial neural network. Initially, for a given reference pixel, first order derivatives in four different directions is calculated. Later maximum (top two) energy variations among calculated derivatives are considered to indicate maximum changes in those specific directions. Further, 3× 3 local reference grid and two 3× 3 local directional grids based on top two maximum magnitude directions are extracted. Finally, relationship among pixels of extracted 3× 3 local grids are encoded using triplet pattern. The complete procedure is named as directional magnitude local triplet pattern (DMLTriP). The retrieval accuracy is measured using two different technique i.e. traditional and learning based CBIR. The parameter like average retrieval precision (ARP) and average retrieval rate (ARR) are considered for image retrieval accuracy measurement on publicly available (natural and medical) image databases. All the experiments discussed in this article clearly shows that proposed feature descriptor outperforms existing state-of-the-art local feature descriptors. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1109/ICIINFS.2018.8721437 | 2018 IEEE 13th International Conference on Industrial and Information Systems (ICIIS) |
Keywords | DocType | ISSN |
Image retrieval,Feature extraction,Medical diagnostic imaging,Image color analysis | Conference | 2164-7011 |
ISBN | Citations | PageRank |
978-1-5386-8492-4 | 0 | 0.34 |
References | Authors | |
0 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Gajanan M. Galshetwar | 1 | 0 | 0.34 |
Prashant W. Patil | 2 | 0 | 0.34 |
Anil Balaji Gonde | 3 | 50 | 5.82 |
L. M. Waghmare | 4 | 18 | 5.65 |
R. P. Maheshwari | 5 | 365 | 11.86 |