|
| 基于编码-解码对称神经网络的高分辨率图像重构机理研究 |
| Research on high-resolution image reconstruction mechanism based on coding-decoding symmetric neural network |
| 投稿时间：2018-09-10 |
| DOI：10.3969/j.issn.1005-5630.2019.04.006 |
| 中文关键词: 卷积神经 反卷积神经 编码 解码 重构 |
| 英文关键词:convolutional nerve deconvolutional nerve coding decoding reconstruction |
| 基金项目:国家自然科学基金资助项目（61405115，61875125）；上海市自然科学基金资助项目（18ZR1425800）；安徽省重点实验室资助项目（CGHBMWSJC03） |
|
| 摘要点击次数: 2710 |
| 全文下载次数: 2384 |
| 中文摘要: |
| 针对目前许多图像重构算法存在重构出来的图像不清晰、分辨率低等问题，提出了一种基于编码-解码对称神经网络的高分辨率图像重构算法。首先将图像进行压缩获取低分辨率图像，然后将低分辨率图像作为输入图像经过编码-解码对称神经网络，并利用其中的卷积神经网络进行编码得到特征图像，最后再利用反卷积神经网络进行解码实现图像的细节恢复。实验结果表明，经过基于编码-解码对称神经网络重构出来的图像比之前的低分辨率图像更加清晰，图像的分辨率得到了提高。 |
| 英文摘要: |
| Aiming at many current image reconstruction algorithms such as JPEG decompression and compressive sensing reconstruction, there are some problems such as unclear image and low resolution. This paper proposes a high-resolution image reconstruction mechanism based on code-decoding symmetric neural network. Firstly, the image is compressed to obtain a low-resolution image, and then the low-resolution image is used as an input image to encode-decode a symmetric neural network, and the convolutional neural network is used to encode the feature image, and finally the deconvolution neural network is used. Decoding implements detail recovery of the image. The experimental results show that the image reconstructed by the code-decoded symmetric neural network is clearer than the previous low-resolution image, indicating that the resolution of the image is improved. |
| HTML 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
