An Efficient Image Retrieval System Using Surf Feature Extraction and Visual Word Grouping Technique

Content Based Image Retrieval is widely used to find the location of images for many application scenarios. It is used to tag the images using large geo-tagged image set. In recent years, images are tagged based on their locations and geo-tagged images consumes more memory space. Therefore, the performance of the image location estimation can be improved by using visual word groups. The mean shift clustering algorithm and a position descriptor have been used to generate visual word groups. A fast indexing structure is build using document builder interface. Thus the drawbacks in the existing system have been over-come in the proposed system. The proposed system involves Speeded-Up Robust Transform (SURF) which is a feature detection and descriptor method which is used for object recognition. The modules include Feature extraction which is used to identify the interest points within the image, Indexing which builds an inverted file structure to reduce the image size, Image retrieval which compares the input image with the query image to give the resultant output and Re-Ranking which categorizes the top ranked images.