Resolution is an important characteristic to determine the nature and features of the image. Enhancing the resolution strengthens the features hidden within the image, and make the image sharper and more informative. The image quality is improved when noise is removed/suppressed from it. The proposed model provides a technique to enhance the resolution of different types of images, obtained from imaging devices, using a convolutional autoencoder. A convolutional neural network (CNN) architecture is developed by adding different layers to the neural network. An autoencoder capable of encoding and decoding the structure of the images is proposed to enhance their resolution. The model tends to learn the lower-dimensional features of unclear images and provide a high resolution to them by predicting and enhancing their dimensions. The model is trained on low-resolution images and the corresponding high-resolution images, and a convolutional auto-encoder is implemented to denoise the image to introduce high-resolution in the blurred or corrupted images. The model overcomes the limitations of the existing denoising filter techniques and provides a higher level of image quality enhancement.
Previous Article in event
A Combined Model-Order Reduction and Deep Learning Approach for Structural Health Monitoring Under Varying Operational and Environmental ConditionsPrevious Article in session
Next Article in event Next Article in session
Image Resolution Enhancement Using Convolutional Autoencoders
Published: 14 November 2020 by MDPI in 7th International Electronic Conference on Sensors and Applications session Applications
Keywords: autoencoder; convolutional neural network; image resolution enhancement