Feature Extraction using Self-Supervised Convolutional Autoencoder for Content based Image Retrieval

Abstract

This paper presents Autoencoder using Convolutional Neural Network for feature extraction in the Content-based Image Retrieval. Two type of layers are in the convolutional autoencoder architecture, they are encoder and decoder layer. The encoder layer extracts the important representation of the image using feature learning capability of the convolutional neural network, and reduces the dimension of the image. The decode layer reconstructs the representation, such that, the output of the autoencoder is close to the input data. The important representation of the image from the encoder layer in convolutional autoencoder, is used as the extracted features in the content-based image retrieval. Similarity distance between the extracted feature of the query image and the database is calculated to retrieve relevant images. The images in Corel dataset are used for the experiment and tested using the proposed model. The experiments show that the extracted features are representable for the images, and can be used to retrieve relevant images in the content-based image retrieval.