Removal Of Random Noise Based On Convolutional Neural Network With Optimization Technique

Abstract

Noise corrupted the digital image because of environmental factor. A new convolution neural network for removing images is presents. The proposed system comprises of convolution neural network with optimization techniques. At first, the digital image is applied into the Batch normalization process for better normalization process. Secondly, the image is given into the Convolution Neural Network (CNN) with Leaky rectified linear unit where the feature is extracted. The CNN process reconstructs the attributes of the recovered image. Finally, to learn and refine the extracted features, the resolved image is introduced in the MSE loss function and the Adam optimization approach. The effectiveness of the proposed framework is tested across a variety of noise levels and compared with various pads and strides. The suggested framework's results show that it outperforms previous denoising algorithms in terms of peak signal-to-noise ratio (PSNR), and it produces the best results.