The image deconvolution eliminates the image blur using the manual methods in which the kernels are applied, and the neural network method uses the optimizers. The optimizers are designed from the examples, or the predetermined manual instructions are given. The inverse problems of the blur images using manual methods have made it difficult for the deconvolution, hence the optimized techniques are designed to get the deconvolution images. Different optimizers are designed to the deconvolution of the images, but the quality of restoration is not as expected. The existing algorithms use the Gradient Descent optimizer for the deconvolution, which is designed for the removal of the Gaussian Noise but not to the other types of noises. The proposed algorithm uses Adam optimizer to remove all the other types of noises like poison and speckle noise. The proposed method is based on deep learning, in which residual convolutional neural network is used for training. The proposed network is trained with Adam optimizer, which increases the Peak Signal to Noise Ratio and Structural Similarity Index Measurement than the previously existing algorithms. The Adam optimizer can also be applied for the real time problems of the image deconvolution.
Image Deconvolution using Deep Learning-based Adam Optimizer
2022-12-01
1406870 byte
Conference paper
Electronic Resource
English
Determining Multi-Frame Blind Deconvolution Resolvability using Deep Learning
British Library Conference Proceedings | 2019
|Image deconvolution using wavelet-based regularization
British Library Online Contents | 2002
|Gradient-based image deconvolution
British Library Online Contents | 2013
|Microgrid Energy Management System with Embedded Deep Learning Forecaster and Combined Optimizer
BASE | 2020
|