基于自编码器的图像去噪设计与实现

Design and Implementation of Image Denoising Based on Autoencoder

文章主要针对自然图像中存在的噪声问题,研究基于神经网络的去噪方法,使之能够更快速地对混合有多种噪音的图像进行有效去噪,从而达到改善图像视觉效果的目的。首先,文章设计一个能够对图像进行编码、解码的神经网络。其次,介绍神经网络的详细结构和网络参数,并针对常见的图像噪声介绍制作训练样本的方法和神经网络的训练流程。最后,展示实验结果。实验表明,文章提出的方法可以有效地去除图像噪声,达到预期效果。

This article mainly focuses on the noise problem in natural images,and researching the denoise methods based on neural network,which makes it possible to effectively denoise images with multiple noises in order to achieve the purpose of improving the visual effect of images. Firstly,we designed a neural network that can encode and decode images. Next,the detailed structure and network parameters of the neural network are introduced,and the methods of making training samples and the training process of the neural network are interpreted for common image noises. At last,experimental results are showed. Our experiment demonstrates that the proposed method can effectively deduct the image noises and reach the anticipated performance.