An improved multi-bags-of-features histograms representation for ear recognition
| Autor: | Hocine Bourouba, Layachi Bennacer, Abdelhani Boukrouche, Hakim Doghmane |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Computer science
business.industry ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Pattern recognition Filter (signal processing) Feature model Support vector machine ComputingMethodologies_PATTERNRECOGNITION Feature (computer vision) Histogram Artificial intelligence Kernel Fisher discriminant analysis Representation (mathematics) business Cluster analysis |
| Zdroj: | 2018 International Conference on Signal, Image, Vision and their Applications (SIVA). |
| DOI: | 10.1109/siva.2018.8660999 |
| Popis: | In this paper, we present a novel approach for ear representation and recognition based on the combined of binarized statistical image feature (BSIF filter), multi-bag of feature model and the overlapping decomposition method to generate the histogram sequence. The recognition accuracy can be enhanced by the following steps. Firstly, the texture information in ear image is extracted using the Binarized Statistical Image Features. Secondly, from the training image responses, the K-means clustering algorithm is used to learn the multi bag-of-feature dictionary. Thirdly, the overlapping decomposition is applied to obtain local ear feature descriptors. Next, the histograms obtained are normalized. Then, the global representation of the ear image is obtained by concatenating all histograms calculated at each level. After that, the discriminant representation of ear image is constructed, using kernel Fisher discriminant analysis. Finally, the k-nearest neighbor and the support vector machine classifiers are used for ear identification. Experiments conducted on IIT-Delhi-1 database; show that the proposed approach provide a significant performance improvement compared to the state of the art in terms of accuracy. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|