The radiation dose of X-ray CT is closely related to the abnormal metabolism of the living body and the induction of diseases such as cancer. At this stage, how to reduce the radiation dose of CT scans and ensure the quality of image reconstruction is a huge challenge for CT image reconstruction technology. The traditional total variation regularization method is not sensitive to the direction of the image. This paper introduces directional gradient information and establishes an adaptive directional total variation sparse angle CT image reconstruction model, which aims to better preserve image edges and improve image reconstruction quality. To adaptively estimate the direction information of the image gradient, based on the geometric moment of the image, an ellipse steerable kernel is used to estimate the main direction of the image edge block. Simulation and actual data results show that the proposed algorithm has good performance in image detail preservation, artifact reduction, and noise suppression.