高级检索
沈卉卉, 李宏伟, 钱坤. 基于RBM无监督学习模型的图像数据去噪[J]. 计算机辅助设计与图形学学报. DOI: 10.3724/SP.J.1089.2023-00050
引用本文: 沈卉卉, 李宏伟, 钱坤. 基于RBM无监督学习模型的图像数据去噪[J]. 计算机辅助设计与图形学学报. DOI: 10.3724/SP.J.1089.2023-00050
HuiHui SHEN, HongWei LI, Kun QIAN. Image Data Denoising Using the unsupervised learning Model of RBM[J]. Journal of Computer-Aided Design & Computer Graphics. DOI: 10.3724/SP.J.1089.2023-00050
Citation: HuiHui SHEN, HongWei LI, Kun QIAN. Image Data Denoising Using the unsupervised learning Model of RBM[J]. Journal of Computer-Aided Design & Computer Graphics. DOI: 10.3724/SP.J.1089.2023-00050

基于RBM无监督学习模型的图像数据去噪

Image Data Denoising Using the unsupervised learning Model of RBM

  • 摘要: 受限Boltzmann机(Restricted Boltzmann Machine, RBM)的深度学习方法在图像去噪中的应用属于未知阶段,它能够无监督学习图像数据的概率分布。基于RBM原理与之前相关算法的研究基础,提出RBM模型的图像数据随机噪声去除的应用。其优势在于它在去噪时无需训练集,也无需预训练,去噪高效且效果好。将原始图像加入随机噪声,把带噪声的图像分割若干小块,将其一一拉成向量,批量输入多个RBM叠加的深度信念网络(Deep Belief Nets, DBN)模型中进行学习,原始图像作为标签进行反向微调,最后将其学习的特征输出,还原成图像,即达到消除随机噪声的目的。用RBM模型算法分别在自然图像数据、模拟的地震数据和真实的地震数据上做随机噪声去除测试试验,试验结果表明,提出的RBM模型在自然图像数据和地震数据上去噪方法可行,且去除随机噪声效果较其他无监督学习算法和卷积神经网络等深度学习方法要好,说明RBM模型在图像特征学习性能、本质特征提取上有很强的能力。也为工程领域中的图像去噪方法提供了一种新的研究思路和借鉴。

     

    Abstract:  The application of restricted Boltzmann machine (RBM) methods in image denoising is still unknown, it can learn the probability distribution of image data unsupervised. Based on the previous research of RBM theory and related algorithms, a new denoising method based on the RBM model can remove the random noise in image data. The advantage is that it does not require training sets or pre-training for denoising. It is efficient and effective in denoising. The original image is added with random noise, and the noisy image is divided into several small blocks, which are pulled into vectors one by one, and then input into the deep belief nets (DBN) with multiple RBM models to learning in batches. The original image is used as a label for fine-tuning, and finally the output is restored to the image, so as to eliminate the random noise. The RBM model algorithm is used to evaluate the effectiveness of removing the random noise in the image data, simulated seismic data and real seismic data. The experimental results show that the proposed denoising method using RBM model is feasible in image data and seismic data, and the effects of removing random noise is better than other unsupervised learning algorithms and some denoising methods with deep learning, such as convolutional neural networks. Furthermore, it shows that the RBM model has a strong performance in image feature learning and essential feature extraction. It also provides a new research idea and reference for the methods of image denoising in the engineering fields.

     

/

返回文章
返回