Autofocus technique is one of the key techniques in digital hologram, which can achieve fast imagine capture and accurate reconstruction. A new wavelet-based autofocus algorithm for terahertz (THz) digital holographic imaging is proposed based on the analysis of focus and defocus characteristics, which mainly concentrates on coefficient of high frequency and low frequency wavelet transform. The new autofocus algorithm of this focus measure is compared with five classic digital hologram autofocus algorithms according to the THz image with and without noise, also comments are attached. The analysis and comparison of the simulation show that the new wavelet-based autofocus algorithm has the following features: a better unimodality, a fineness sharpness and a stronger performance of anti-noise.