Image Classification Algorithm based on Multi- Feature Extraction and KNN Classifier
Keywords:
KNN, local binary patterns, colour moment, canny edge detection, feature extraction, image classificationAbstract
The classification is task of data mining where the algorithms are first trained on the patterns and then
classified according to their past experience. The proposed image classification technique works in similar fashion, thus
first the set of well-defined images with their classes are used to train the model and after training the model is used to
identify the classes of the test set.In order to train and test the model multiple low-level features of image such as texture,
colour and edges are used .The proposed algorithm uses local binary pattern for texture featureextraction, canny edge
descriptor for edge detection and colour moment for colour feature extraction. These features are used with the KNN (knearest classifier) for finding the similarity of the domain specific images.The experiment results of proposed work show
that multiple feature extraction techniques improve classification performance as compared to single local binary
pattern feature extraction technique.
