Computed Tomography (CT) imaging technique is widely used in geological exploration, medical diagnosis and other fields. In practice, however, the resolution of CT image is usually limited by scanning devices and great expense. Super resolution (SR) methods based on deep learning have achieved surprising performance in two-dimensional (2D) images. Unfortunately, there are few effective SR algorithms for three-dimensional (3D) images. In this paper, we proposed a novel network named as three-dimensional super resolution convolutional neural network (3DSRCNN) to realize voxel super resolution for CT images. To solve the practical problems in training process such as slow convergence of network training, insufficient memory, etc., we utilized adjustable learning rate, residual-learning, gradient clipping, momentum stochastic gradient descent (SGD) strategies to optimize training procedure. In addition, we have explored the empirical guidelines to set appropriate number of layers of network and how to use residual learning strategy. Additionally, previous learning-based algorithms need to separately train for different scale factors for reconstruction, yet our single model can complete the multi-scale SR. At last, our method has better performance in terms of PSNR, SSIM and efficiency compared with conventional methods.scanning electron microscopy (SEM) image. Li proposed a voxel SR reconstruction algorithm 11 based on sparse representation, which can improve the resolution in all directions.Zhang et al. extended adjusted anchored neighborhood regression algorithm (A+) 14 , to 3D and proposed high frequency modified 3DA+ algorithm 15 , where a correlative dictionary and mapping matrix between high frequency and low frequency was established. In reconstruction stage, the matched dictionary atom and mapping matrix were searched for each input of the 3D block to complete SR.Unfortunately, the aforementioned algorithms are focused on 2D images, in view of the fact of 3D-CT images of rock, the following issues remain to be solved: First, the computational intensity and memory of 3D image data is far greater than the 2D images, so the method to handle with 2D images can't be directly transferred to 3D model; Second, CT samples are not as convenient as 2Dimages to obtain, that is to say, it's not easy to get substantial alignments of rock CT samples to training network. In addition, CT image of rock has the characteristics of low contrast, single texture, and complex pore structure, which all bring difficulty to task of SR; Third, during training network and reconstruction stage, the calculation and time complexity have to be taken account to ensure our work can be carried out on the general computing equipment. Hence, it is desirable to devise a new network to cope with SR for voxel images.In order to enhance resolution of CT images of rock from three directions (i.e., x, y ,z), we propose a novel network, termed as 3D super-resolution convolutional neural network (3DSRCNN), to promote resolution for volumetric images. Bef...