To reconstruct the image of a deep space target by using non-uniform sparse launch array of a Fourier Telescopy(FT), a new method based on the least square fitting was proposed to estimate the lack of Fourier component of the target. Firstly, the T-sparse emitter array of laser transmitter system was used as transmitter configuration of FT, and the returned time-domain signal was processed by a DC filter. According to the basic principle of FT, the signal was demodulated to obtain a triple product by the phase closure. The least square fitting method was used to estimate and fit the absent Fourier components and to take it for the information recovery basis. Finally, the image was reconstructed with inverse non-uniform Fourier transform. Four targets were simulated with different SNRs. As compared with the original method, the new method resolutes the details of targets and the Strehl Ratios are improved by a maximum value of 0.074 2 and the minimum value of 0.009 8 at a SNR of 200 db and seven rank fitting. Obtain data in a field experiment show that the new method overcomes the distortion of reconstructed image come from frequency spectral deviation and provides a reference for practical engineering.