Abstract: In order to improve analyzing and processing capability for time-varying signals, Q-shift filters meeting no aliasing, perfect reconstruction and orthogonality are designed in this paper. Compared to the traditional Q-shift filters, coefficients of the proposed ones can be adaptively computed according to different characteristic of the input signals. While applying the adaptive Q-shift dual tree complex wavelet based threshold denoising for ultrasonic flaw signals, the experimental results show that the signal-to-noise of the original signals can be enhanced greatly.