以鄂尔多斯盆地某区块的砂泥夹层岩心为研究对象，使用基于小波变换的去噪神经网络（denoising neural network based on wavelet transformation, DWTNet）对于岩心的图像进行去噪研究。该方法的评断结果采用峰值信噪比（peak signal to noise ratio, PSNR）和去噪后的图像结果进行了对比。研究表明，利用DWTNet在测试集YX1、YX2测试所提出的算法，并与EGDNet等去噪算法进行对比，PSNR在噪声为25、50、75 dB时，高于EGDNet算法0.527、0.418、1.1 dB。所提的算法在峰值信噪比等指标均高于其他算法；并在视觉效果上其处理得到的图像也更加清晰。方法的提出对于孔隙度、平均体积比表面积、平均曲率计算等都有着非常重要的意义。

Based on the sand-mud interlayer core of a block in Ordos Basin, denoising neural network based on wavelet transformation (DWTNet) was used to denoise the core image. The evaluation of this method was carried out by comparing the peak signal-tonoise ratio (PSNR) and the post-denoising image outcomes. The investigation reveals that by applying the DWTNet denoising algorithm to the test sets YX1 and YX2, and contrasting it with other denoising algorithms such as EGDNet, the PSNR values at noise levels of 25, 50, and 75 dB are respectively 0.527, 0.418, and 1.1 dB higher than those achieved by the EGDNet algorithm. The proposed algorithm surpasses others in terms of metrics including peak signal to noise ratio(PSNR), and visually, the resulting images processed by it exhibit enhanced clarity. The introduction of this method holds substantial significance for the calculation of parameters like porosity, mean specific surface area, mean curvature, among other rock properties, thereby advancing the capabilities in digital core technology, CT scanning analysis, and understanding of rock characteristics.