Image Mining by Multiple Features
2020
This paper discusses an image mining algorithm for efficient and accurate image retrieval. The proposed method uses multiple image features combination. Color feature in the form of first-order, second-order and third-order moment is calculated on a sub-block. These moments calculate statistics features over a local region. The computed first-order moment is considered as mean. Standard deviation and skewness are considered as second-order and third-order moment, respectively. These features are calculated for images and stored in database. For texture extraction, center symmetric local binary pattern (CSLBP) feature is used. CSLBP is straightforward in computation and it is also rotation consistent. CSLBP generates histogram of 16 bins for one sub-block. In order to capture minute details around local region, Laplacian of Gaussian (LoG) is used as a weight factor during CSLBP histogram construction. This weight factor makes texture feature more powerful and increases its discrimination power. Resultant weighted histogram is quantized and binary result is stored in the database. For test image, its texture and color features are compared with the stored texture and color features. Based on the color and texture feature’s correlation with test image, results are retrieved. Our results clearly show that by incorporating multiple features as well as local weight factor, the proposed image mining gives desirable retrieval results.
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
21
References
0
Citations
NaN
KQI