Abstract: Radon Fourier transform (RFT) based on ChirpZ transform (CZT) is a fast RFT which is implemented in frequency domain. However, it brings about large computation for traversing Doppler ambiguous numbers, especially when the objects are seriously ambiguous in Doppler domain. Moreover, the procedure of maximum selection among the accumulation results on different Doppler ambiguous numbers also increases the noise power. To solve that, a Coarse to Fine RFT (CTF-RFT) algorithm is proposed. In the coarse detection stage, the Doppler ambiguous numbers are coarsely searched instead of traversal search as the traditional method did. Once the accumulation results on several Doppler ambiguous numbers are obtained, potential targets are extracted and are lined along the ambiguous number dimension to further eliminate false targets according to the law of amplitude variation of the true targets in the ambiguous number dimension. In the fine detection stage, the final parameters of the targets are determined in the Doppler ambiguous range which is determined by the coarse detection stage. Compared with CZT-RFT, the proposed CTF-RFT reduces the computational complexity obviously while keeping its detection performance almost unaffected. Theoretical analysis and experimental results prove the validity of the algorithm.