The proposed model has been designed by combining the various feature descriptors, which includes the histogram of oriented gradients (HoG), Tamura and scale invariant feature transform (SIFT), for the purpose of feature extraction from the images in given database. The feature extraction process is followed by the classification model based upon the support vector machine (SVM). The SVM classification is a probabilistic classification algorithm and utilizes the concept of support vectors to prioritize the strong features for the matching of the testing samples with the training database to shortlist the matching samples on top of the index. The proposed model has undergone the various experimental iterations, where the performance of this model is analyzed upon the basis of various parameters such as feature extraction delay, classification delay, accuracy, precision and recall. The proposed model has been found improved on the basis of all of the performance parameters except feature extraction delay, because it involves all features (feature descriptor ensemble) to classify the images under CBIR model. The proposed model has been discovered to be improved by nearly 3-5% on the basis of all of the performance parameters in comparison to the other CBIR model.
CBIR, Histogram of Oriented Gradients, SIFT, SVM.