Abstract: 3D image mosaic is a crucial technology to obtain the whole high-resolution image of large-size object by cone beam computed tomography(CBCT) system,which has become a new research direction of 3D image processing.Aiming at the low matching accuracy of similar features and the time-consuming of feature matching process,we propose a fast feature matching algorithm based on global binary descriptor.The method firstly extends the original BRIEF(Binary robust independent elementary features descriptor) to the 3D case to suit 3D images.Then a global binary descriptor is added to it to enhance the discrimination of the 3D extended feature descriptor.Finally,according to the characteristics of feature descriptor mentioned above,a coarse-to-fine feature matching strategy is designed to improve the accuracy and efficiency of feature matching.Experiments show that the proposed method is simple and effective,which can bring down the number of mismatches and improve the matching speed obviously.