With unique properties, multispectral cameras have become a hot research direction in recent years. In our country's major space mission "Mars Exploration Project", the multispectral camera was mounted on the "Zhu Rong"rover as an important payload to complete a number of exploration missions. Due to the optical structure of the multispectral camera and other reasons, there are often deviations such as rotation, scale change, and displacement between the images of each channel. For multispectral images, there may be huge differences between the image grayscales of different channels. For the same target subject, the local grayscale contrast may even be opposite. Therefore, the conventional image registration method is difficult to solve the alignment issue between the subjects in each channel. In this paper, a calibration method based on a pyramid mixture model of the circular templet is proposed, and a set of accurate calibration parameters is obtained by using the templet images, so as to complete the channel registration of the Mars multispectral image. At the same time, based on the particularity of the template images, an objective evaluation criterion of geometric calibration is proposed. © COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.