Micro-computed tomography is widely used for medical diagnosis and non-destructive testing since the 1970s. In three-dimensional micro-CT imaging, the quality of the reconstructed images depends highly on the precision of the geometric parameters of the system. Image quality could be severely degraded by misaligned parameters, resulting in artifacts, blur and lower resolution in images, especially for CCD-coupled CT system (CCT system). Therefore, high-precision methods for estimating CT system parameters are necessary. Different research groups have proposed various methods to estimate geometric parameters for fan-beam CT or cone-beam CT. In terms of data acquisition methods, the main methods can be categorized as multi-angles projection and single-angle projection. However, due to the narrow field of view caused by the ultra-high resolution of the CCD-coupled detector, neither of these two methods can be applied to CCT system directly. Generally, a CCT system consists of an X-ray source, a CCD-coupled detector, and a precision rotation stage for placing samples, in which the X-ray source and detector are distributed on both sides of the sample. During scanning, the sample rotates around the rotation axis, while the X-ray source and the detector remain stationary. In an ideal CCT system, strict geometric alignment must be satisfied. It can be summarized as follows: Firstly, the X-ray focal spot, the axis of the rotation, and the center of the detector should be in a plane; secondly, the straight line which passes the focal spot and the detector center, should be perpendicular to the axis of rotation and the detector; thirdly, the axis of the rotation should be parallel to the row of the detector. Basing on precious studies, there are five geometric parameters to be calibrated totally in a CCT system. We conclude all the calibrated parameters as follows: 1) u0,v0 are projection coordinate of the X-ray source spot on the detector; 2) η is the angle of detector tilt around X axis; 3) R is the distance between X-ray source and rotation axis; 4) D is the distance from X-ray source to detector.In this paper, a mathematical model based on a double micro-sphere phantom is proposed to estimate all the above geometric parameters of CCT system. The double micro-sphere phantom consists of two spheres with a diameter of 100μm. Generally speaking, when a sphere rotates around the rotation axis, the centers of its projections on the detector form an ellipse or a line. To estimate all the parameters, the calibration is carried out in two steps. Firstly, the circular scanning imaging of the double micro-sphere phantom is carried out to obtain its elliptical projection trajectory. Then the mathematical relationship between the ellipse equation corresponding to the trajectory and the geometric pose parameters of CCT system is established to solve the geometric pose parameters.To verify the effect of the calibration method in realistic imaging system, an experiment was performed in a misaligned CCT system. After the parameters calibrated using the mathematical model proposed in this paper, a bamboo fiber was scanned by the CCT system. Then, the results of the calibration parameters were substituted into the reconstruction algorithm, and the reconstructed images were compared before and after parameters calibration. Experimental results show that by substituting the corrected geometric parameters solved by the mathematical model into the reconstruction algorithm, the aliasing and blurring of edge information in the reconstructed images are reduced yielding an improved image detail resolution. In addition, we adopted the Energy of Gradient (EOG) to quantitatively evaluate the image quality. The EOG value of reconstructed image after calibration is larger than before, which also shows the image quality after calibration is improved.This paper proposes a new method to estimate geometric parameters of CCT system. The method allows us to estimate all the parameters by acquiring projections of the designed double micro-spheres phantom. Compared with other methods, it could be used to CCT system which has high resolution and narrow field of view. Meanwhile, this method removes the influence of rotation errors on geometric pose parameters. The accuracy of this method is confirmed by experimental results. At last, a bamboo fiber sample was scanned and reconstructed. The experimental results show that the artifacts are reduced and image quality is improved.