Wavelet estimation is the foundation of high-resolution seismic data processing and interpretation. It is of great significance to estimate the wavelet rapidly and exactly. This paper gets the amplitude spectrum from the trispetrum and phase from the seismic data based on the principle of kurtosis maximum, which avoids phase wrapping properly, and discusses the application condition of this principle. The property of the different window functions is also compared, and Hamming is moved to calculate the time-varying wavelets. The ricker wavlet model is used which is frequency-changing to illustrate its effectiveness. In the real data processing, the method is comparatively analyzed for the well-seismic calibrated data based on different frequency ricker wavlets. The results show that the estimated wavlet amplitude spectrum is more stable, and the phase spectrum can be realized easily and is more consistent with the true wavelet.