Abstract: In the single channel circular synthetic aperture radar (SAR) system, the resolution in altitude dimension is limited severely. To solve this problem, a three dimensional(3D) imaging algorithm based on the joint sparse constraint, which consists of the two-dimensional(2D) processing followed by the separated imaging in the elevation dimension, is proposed in this paper. Firstly, the phase error introduced by the mismatched height in traditional 2D imaging is analyzed and deduced. Then, the mismatched phase errors of imaging grids at different locations are calculated to build an over-complete dictionary, and the spectrum of the processed radar data is projected on the dictionary based on the joint sparse constraint. Since the coherence relationships between the phase error of targets and the vectors in the dictionary are different, most of the energy of the spectrum will be projected on the positions corresponding with the matched height in the dictionary. The joint sparse constraint can realize the aggregation of projection energy and the suppression of cross components, thus implementing the separation of mixed received data and focus processing of targets with different heights. Finally, numerical simulation verifies the effectiveness of the algorithm.