An improved precoding scheme for multiple-input multiple-output (MIMO) broadcast channel (BC) is proposed to maximize the detection signal-to-interference-plus-noise ratio (SINR) at user side based on the vector-perturbation technique. With the derived maximum detection SINR criterion, a new tree search based detection algorithm, called iterative M-algorithm (IMA), is utilized to find out the optimal perturbation vector. Simulations show that the proposed scheme outperforms the existing schemes which are based on the maximum detection signal-to-noise ratio (SNR) criterion. Moreover, the proposed scheme can achieve the same bit error rate (BER) performance as the vector-perturbation scheme based on sphere encoder and the maximum detection SINR criterion, with guaranteed polynomial worst-case complexity. An improved precoding scheme for multiple-input multiple-output (MIMO) broadcast channel (BC) is proposed to maximize the detection signal-to-interference-plus-noise ratio (SINR) at user side based on the vector- perturbation technique. With the derived maximum detection SINR criterion, a new tree search based detection algorithm, called iterative M-algorithm (IMA), is utilized to find out the optimal perturbation vector. Simulations show that the proposed scheme outperforms the existing schemes which are based on the maximum the proposed scheme can achieve the same bit error rate (BER) performance as the vector-perturbation scheme based on the sphere encoder and the maximum detection SINR criterion, with guaranteed polynomial worst- case complexity.