Feature Selection based Reduction in Dimensions and Indexing of Images for Efficient Image Retrieval

• Published in: 2021 International Conference on Emerging Smart Computing and Informatics (ESCI) pp. 456 - 461
• Main Authors: Prasanthi, B, Pabboju, Suresh, Devara, Vasumathi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05.03.2021
• Subjects: Content-based Image Retrieval; Dimension Reduction; Feature extraction; Features Extraction; Image retrieval; Indexing; Informatics; Model Semi-supervised Learning; Prototypes; Query Image; Tagging
• DOI: 10.1109/ESCI50559.2021.9397054

CBIR prototype recovers images from massive databases identified by the interrogatory images provided by the client. The stocks of CBIR prototypes extract the limited amount of feature variable sets that impound the retrieval adequacy. CBIR is chiefly used in the domain of data retrieval where it is not necessary to feed information in order to search for a specific image stored in the database. Instead, feeding an image that akin to the one that is being traced would suffice. This proposed work eliminates the traditional approach of storing the images in the database which is tagging the images as per their content and retrieving them using those keywords. Content-based image recovery (CBIR) is attractive because searches that depend simply on metadata are subject to annotation aspect and fulfillment. The proposed framework efficiently performs most similar image retrieval from the database corresponding to the test query image. Our approach outturns state-of-the-art attainment on the Corel 10K dataset while being much fast at particular to test time. The results obtained were higher-up to other state-of-the-art CBIR prototypes analogous to precision.