An Ensemble of Feature Extraction with Whale Optimization Algorithm for Content Based Image Retrieval System

In recent days, content based image retrieval (CBIR) becomes a hot research area, which aims to determine the relevant images to the query image (QI) from the available large sized database. This paper presents an optimal hybrid feature extraction with similarity measure (OHFE-SM) for CBIR. Initially, histogram equalization of images takes place as a preprocessing step. Then, texture, shape and color features are extracted. The texture features include Gray Level Co-Occurrence Matrix (GLCM) and Gray Level Run Length Matrix (GLRLM) is extracted, where the optimal number of features will be chosen by whale optimization algorithm (WOA). Afterwards, the shape feature extraction takes place by Crest lines and color feature extraction process will be carried out using Quaternion moments. Finally, Euclidean distance will be applied as a similarity measure to determine the distance among the feature vectors exist in the database and QI. The images with higher similarity index will be considered as relevant images and is retrieved from the database. A detailed experimental validation takes place against Corel10K dataset. The simulation results showed that the proposed OHFE-SM model has outperformed the existing methods with the higher average precision of 0.915 and recall of 0.780.