A New Content-Based Image Classification Method Using SVM-weight and Euclidean Distance

摘要

As the demands and usages for image classification and retrieval increase, the classification methods need to be improved to be more automatic and effective. The traditional text-based image classification fails to recognize the underlying content of image that would lead to misclassifying issue. In this paper, a new classification based on weighted Euclidean distance whereas the weights are estimated via Support Vector Machine (SVM) is proposed. To overcome the problem of misclassification and increase the classifier accuracy for some particular classes, the new classification method uses the weight set estimated from SVM, which is then applied to the Euclidean-based K-Nearest Neighbour (K-NN) method. The experiments have revealed the proposed method has an accuracy of 93.75%, which is better than traditional classification such as K-NN and SVM. Moreover, the accuracy of the image feature vector has minimal impact on the accuracy of the method compare to other classifier. The SVM weighted K-NN classifier has provided a promising direction for criteria-based image retrieval.