Content Based Image Retrieval System Using Integrated ML and DL-CNN

Abstract

In this article, content-basedimage retrieval (CBIR) system is developed since it has significant scope for research in image processing domain. In CBIR, visual contents are being used to search an image from a huge scale image database as per users interests and based on automatically derived query image features. The term 'content' might refer to low level features like color, shape or texture extracted from the image. Several research challenges can be addressed towards the design and development of CBIR systems, very few techniques are available to address and solve the problem of semantic gap presented in images which are not efficient.Themachine learning (ML) method has investigated as a practicableapproach to decrease the semantic gap. Also motivatedfrommodernfulfillment of deep learning for image processing applications, we focused to deal with an artificial intelligence based deep learning approach, treated as Convolutional-Neural-Network (CNN), for the purpose ofsimilarity measurement of accuratesemantic features. In this article, the usage of CNN for the image retrieval issues is investigatedwith their solutions using deep learning approaches.Further, it is also incorporated with principal component analysis (PCA) for extracting salient features from the images. Euclidean distance measurement is employed for the similarity evaluation of extracted feature vectors of query image and database images. Extensive simulation results on different image categories discloses that proposed DL-CNN-ML outperforms existing CBIR approaches like ML and CNN in terms of mean average precision (AP), mean average recall (mAR) and F-score values.