Abstract: Channel estimation is obtained by processing the received signal of the base station (BS) and the known pilot sequence sent by the user. Two-dimensional (2D) nested arrays save system cost and enable performance of massive MIMO (multiple-input multiple-output). However, the structure of 2D nested arrays is not a regular planar arrays, so it's difficult to directly process received signal from BS. In this paper, a 2D-DFT-based channel reconstruction algorithm was proposed. Firstly, the received signal was autocorrelation processed and converted into received signal of the continuous differential co-array. Secondly 2D-DFT was used to estimate the initial DOA of each path of the user, and the angle rotation technique was used to enhance the DOA estimation to achieve super-resolution estimation. Then channel gain was estimated by conventional LS estimation method based on accurate DOA estimation. Finally, user's channel was reconstructed. Various numerical results are provided to corroborate the proposed studies.