Please use this identifier to cite or link to this item:
Title: Searching Technique for Near Exact Duplicate Images in Cloud Database
Authors: Maneesha
Supervisor: Chana, Inderveer
Keywords: Cloud computing;Duplicate Images
Issue Date: 9-Aug-2016
Abstract: Present technologies have made image capturing cameras available to everyone and easy access to internet. Internet based services and high popularity of the social networking sites that are used for the sharing the picture has resulted in a large quantity of images shared on the internet. The availability of the cloud storage allowed the individual and companies to keep and maintain the huge collection of images. Some of them are the near exact duplicate images. Near exact duplicate images are the images which are generated after applying some modification such as cropping, rotation etc. These images are stored with the original images, so one image is stored multiple times. Traditional database are not fit for these types of images because they are unable to access the information within the images. These were intended to deal with text based structure and store the information about image like metadata, resolution. They store the image in files but they cannot perform image processing, classification and image matching based on the information stored in images. In this thesis techniques of the feature detection, feature descriptor and nearest neighbour are discussed. The system is designed for implementation of searching near exact duplicate images in cloud database. MATLAB 2015a is used for implementation of proposed solution. Here Binary Robust Invariant Scalable Keypoints (BRISK) is used as a feature descriptor which is a binary descriptor and for indexing Locality sensitive hashing is used. This allows the user to issue a query image and then search all the near exact duplicate images related to query image. Experiment results show that it effectively searches near exact duplicate images. The performance of the proposed solution is discussed by using precision and recall.
Description: ME Thesis
Appears in Collections:Masters Theses@CSED

Files in This Item:
File Description SizeFormat 
4051.pdf2.74 MBAdobe PDFThumbnail

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.