Content Based Image Retrieval Based On Color And Texture Features Using KNN Classifier

Abstract

Content based image retrieval (CBIR) is a technique to retrieve desired images from large databases against a query image on the basis of some similarity between features extracted from images itself. Here content based defines that search is on the basis of contents of the image rather than text based search using keywords, or some tags or metadata linked with the image. CBIR uses the information present in an image for query formulation and to give result. As the internet grows exponentially, it leads to huge number of digital images embedded in databases and traditional text based searching becomes tedious. A fast and efficient search technique required to deal with images in large volume which leads to this new field. In this thesis a new method is proposed which use color and texture features of image for construction of feature vector. K-Means clustering is used to cluster similar type of images and finally KNN is used to retrieve the desired results.