Relevance Feedback Based CBIR System Using SVM and Bayes Classifier
Navneet Kaur, Sonika Jindal, Bhavneet Kaur · 2016
Image search techniques were not generally basedon visual features but on the textual annotation of images. Images were firstly annotated with text and then searched usinga text-based approach from traditional database managementsystems which is time consuming and difficult to manage. Toovercome this problem, CBIR (Content Based Image Retrieval) is introduced which is becoming the hottest research areathese days due to vast range of real time applications suchas Crime Prevention, Photograph Archives, Medical Diagnosis, Geographical Information and Remote Sensing System etc. TheCBIR system consist of various phases to extract and matchthe features and search the images from the large scale imagedatabases on the basis of visual contents such as Color, Shape andTexture according to the user's interest. As Semantic Gap is themost important and challenging issue. In this paper, RelevanceFeedback is used to deal with this issue which based on SupportVector machine has been extensively used in the CBIR systemto bridge the semantic gap between low level features and highlevel human perception features. The learning techniques arepredominently used for the classification of images in lablelledand unlabelled datasets. In our proposed work we have to workon KNN, SVM and Bayes Classifier to classify the images. The implementation of our proposed work is done in OpenCvand experiments conducted on the Corel Dataset having 10,000images. After attempting the experiments on various images wehave to calculate the Precision and Recall which represent in theform of graphs. After analyzing the results we have concludedthat our method is effective to reduce the semantic gap.