Digital Rock Core Images Super Resolution via SRCNN Based on Accelerated Bicubic Interpolation
Digital Rock Core Images Super Resolution via SRCNN Based on Accelerated Bicubic Interpolation
复制标题
基于加速双三次插值的 SRCNN 数字岩心图像超分辨率
DOI：
10.1145/3411016.3411162
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
V. Popov
中科院分区：
文献类型：
--
作者：
Yunfeng Bai;V. Berezovsky;V. Popov
超分辨率卷积神经网络（SRCNN）的能力已被证明可以提高图像的分辨率。将SRCNN应用于岩心数字图像的增强，对岩心分析具有重要作用。在这个过程中，我们注意到双三次插值算法是SRCNN的第一步，可以通过调整计算策略来提高速度。我们提出了一种基于加速双三次插值的SRCNN，并使用2000幅数字岩心图像进行了性能测试。实验结果表明，该算法比改进的基于区域的双三次图像插值算法和标准双三次图像插值算法具有更快的插值速度，验证了基于该算法的SRCNN生成更高分辨率数字岩心图像的可行性。
The capability of Super Resolution Convolutional Neural Networks (SRCNN) has been proved to enhance resolution of images. We applied SRCNN to enhance digital rock core images that play an important role in analyzing rock core. In this process, we noticed that the bicubic interpolation algorithm that is the first step of SRCNN might improve the speed by adjusting the calculation strategy. We proposed an SRCNN based on accelerated bicubic interpolation and tested the performance with 2000 digital rock core images. The experiment demonstrated that the accelerated bicubic interpolation algorithm faster than improved region-based bicubic image interpolation algorithm and standard bicubic interpolation algorithm, and demonstrated the feasibility of SRCNN based on our proposed algorithm to produce higher resolution digital rock core images.