Content-Based Image Retrieval Using Multilevel Robust Mechanism

Abstract

Abstract Content-Based Image Retrieval only concentrates on the information about the image like its texture, color, shape, homogeneity, and contrast rather than the image tags, description, and keywords. Traditionally many machine learning algorithms are used to retrieve the information but fail to acquire accuracy and created overfitting problems. Not only that large image database handling becomes quite hard to achieve quicker results without human intervention. Hence, we are opting for a robust mechanism for every step of retrieval. Image pre-processing inhibited AlexNet architecture with latent feature extraction and feature selection wear out Mutual Information neural estimation, which is used to handle high dimensional images with fixed resolution and removes the redundancy of images in the database. The similarity measure using Euclidean distance is employed to retrieve the images. The accuracy and performance efficiency is considerably high and pertinent.