Based on the properties of Mojette projection transform in spatial and frequency domains, we propose a limited-angle computed tomography (CT)reconstruction algorithm with minimum redundancy coverage in the frequency domain. The minimum redundant coverage corresponds to the minimum spatial projection sampling (i.e., the reconstructed image has minimum number of projections with maximum efficiency). The equivalent relationship of the Mojette projection data in its frequency domain is determined. The limited-angle CT image reconstruction can be achieved by compressing the projections into a limited angular range. Experimental results show that high-quality reconstructed images can be obtained by using the proposed algorithm.