An Novel Framework For Content Based Image Retrieval With Quality Assessment System using Optimal Deep Convolution Neural Network
P. Anandababu1, M. Kamarasan2

1P. Anandababu, Research Scholar, Department of Computer and Information Science, Annamalai University, Chennai.
2M. Kamarasan, Assistant professor, Department of Computer and Information Science, Annamalai University, Chennai.

Manuscript received on September 16, 2019. | Revised Manuscript received on 24 September, 2019. | Manuscript published on October 10, 2019. | PP: 3916-3926 | Volume-8 Issue-12, October 2019. | Retrieval Number: L34281081219/2019©BEIESP | DOI: 10.35940/ijitee.L3428.1081219
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (

Abstract: Content based image retrieval (CBIR) models become popular for retrieving images connected to the query image (QI) from massive dataset. Feature extraction process in CBIR plays a vital role as it affects the system’s performance. This paper is focused on the design of deep learning (DL) model for feature extraction based CBIR model. The presented model utilizes a ResNet50 with co-occurrence matrix (RCM) model for CBIR. Here, the ResNet50 model is applied for feature extraction of the QI. Then, the extracted features are placed in the feature repository as a feature vector. The RCM model computes the feature vector for every input image and compares it with the features present in the repository. Then, the images with maximum resemblance will be retrieved from the dataset. In addition, the resemblance between the feature vectors is determined by the use of co-occurrence matrix subtraction process. Besides, structural similarity (SSIM) measure is applied for the validation of the similarity among the images. A comprehensive results analysis takes place by the use of Corel 10K dataset. The experimental outcome indicated the superiority of the RCM model with respect to precision, recall and SSIM.
Keywords: Deep learning, CBIR, Feature Extraction, ResNet
Scope of the Article: Deep learning