The early fault feature information of rolling bearing is very weak and difficult to extract, and the selection of parameters of the tunable Q-factor wavelet transform (TQWT) depend on the user's experience. The method of early fault diagnosis of rolling bearing based on improved TQWT is proposed to solve the above problems. Firstly, the range of Q-factor is presented. The several components are obtained by decomposing the fault bearing acceleration vibration signal using TQWT. Then, the each component is demodulated using the envelope derivative energy operator. The best decomposition parameters is selected adaptively according to the index of characteristic frequency strength factor in the energy spectrum. The optimal decomposition results of the fault signal can be obtained. Finally, the fault feature information can be accurately extracted by analyzing the envelope derivative energy spectrum of the optimal component. The method is verified by analyzing simulation signal, experiment data and engineering case. The results show that the method can effectively extract the early weak fault characteristics of rolling bearing and accurately judge the type of fault. It has a certain value of engineering application.