Accurate pore structure characterization, as a basic tool for efficient exploration and development in reservoirs and digital rock, has become increasingly popular nowadays. However, using single-scale digital core data, it is difficult to evaluate the multi-scale pore structures in shales. This study proposes an integrated workflow for identifying and extracting pore parameters from multi-scale three-dimensional and two-dimensional digital rock images, which includes full-diameter core computed tomography (CT), micro-CT, focused ion beam-scanning electron microscopy and scanning electron microscopy images. This workflow realizes the identification and parameter extraction of pores and fractures from mesoscopic to microscopic scales. First, meso-fractures are extracted using the connected domain analysis method from full-diameter CT images, and the apparent attitudes are calculated using the least squares and connected domain analysis method. Then, micropores and fractures are identified from Micro-CT and focused ion beam-scanning electron microscopy data, and the pore network models are established. Features, including pore radius, surface area, volume, throat radius, length, and coordination number, are calculated based on the maximum ball method. Different types of pores in scanning electron microscopy images are automatically identified using deep learning methods, and the pore parameters are computed using connected domain analysis methods. Subsequently, the workflow is applied to a practical case and the results show accurate extractions of pore structure information. This study provides important guidance and support for the quantitative evaluation of pores and fractures in unconventional reservoirs.