Content based word image retrieval using multi-layer perception based convolutional neural networks

Content based image retrieval (CBIR) plays the major role in real time applications likes search engines, libraries. The conventional CBIR systems are implemented by using basic image processing and machine learning models. So, they resulted in the poor performance against various situations. Therefore, this article is focused on implementation of CBIR system using multi-layer perception based convolutional neural network (MLP-CNN). Initially, principal component analysis (PCA) applied on images to perform the dimensionality reduction operation, which also extracts the content specific features. Further, MLP-CNN model is used to train the system and generates the trained features. Finally, the testing operation is performed using MLP-CNN, which generates the output as content specific images. The simulation results shows that the proposed method resulted in superior performance as compared to state of art approaches.