SIM techniques based on synchronous phase shift and vertical scanning have been proposed for fast and precise three-dimensional (3D) measurement. These methods are also named as time-domain phase shift (TSIM) method. In these methods, the envelope of light intensity depth response (IDR) curve is extracted to achieve the 3D map. However, when this method is applied in rough surface or complex environment, the imaging noise and background intensity variation will severely affect the shape of IDR which brings difficulties in envelope detection and causes large measurement errors. In this paper, we propose an improved TSIM (ITSIM) technique to enhance the measurement accuracy and robustness of TSIM. In this technique, the modulation distribution of each image is first calculated by a global frequency analysis algorithm, and the modulation depth response curve (MDR) of each pixel can be obtained. These MDR curves are then multiplexed with the corresponding IDR curves to correct the shape of the IDR curves. Finally, the envelope of the corrected IDR curve is detected to accurately reconstruct the 3D shape. The proposed method can effectively resist the intensity fluctuation and imaging noise, and thus can improve the measurement accuracy compared with the conventional TSIM. The experiments performed on standard step sample show that the root mean square error of TSIM and ITSIM is 8.5nm and 21.2nm, respectively, justifying the feasibility of proposed ITSIM.