SCRS Conference Proceedings on Intelligent Systems
A Graphical Approach for Image Retrieval Based on Five Layered CNNs Model
Authors: Mohammad Khalid Imam Rahmani
Publishing Date: 25-04-2022
ISBN: 978-93-91842-08-6
Abstract
Image processing is an important field in the computer vision domain. A lot of work has been done for the processing of image data in various fieldslike science and technology, defense, medical, space science for satellite imagery analysis, seismology, traffic control, crime control,publishing, and other emerging research areas.There are different levels of complexities for the accurate retrieval of images as most of the images are affected by different kinds of noise and other factors. In this proposed work, I have performed the work of image retrieval using two methods: firstly, processing for denoising and filtering of the dataset of images taking density parameter 0.7 and adaptive gamma parameter constant value 0.5. The obtained images are then processed by Convolutional neural networks (CNNs). The 5-layer convolutional neural network has been used for the best features extraction and then the algorithm is finally optimized using GA (Genetic Algorithm). In my work I have used 5*5 fold convolutional layers and compared the results with the previous approach Deep Convolutional neural network (DCNN). Finally, the Genetic Algorithm is implemented to obtain the best-optimized value. The proposed work is validated with a graphical-based approach using the mathematical results in terms of peak signal-to-noise ratio (PSNR), mean-squared error (MSE), and the processing time of the algorithm. The result parameters of the proposed algorithm clearly show better performance as compared to the previous approach.
Keywords
PSNR, MSE, GA, CNNs, Image Processing, Image Retrieval.
Cite as
Mohammad Khalid Imam Rahmani, "A Graphical Approach for Image Retrieval Based on Five Layered CNNs Model", In: Raju Pal and Praveen Kumar Shukla (eds), SCRS Conference Proceedings on Intelligent Systems, SCRS, India, 2022, pp. 261-270. https://doi.org/10.52458/978-93-91842-08-6-26