An improved semi-fragile algorithm incorporated the generalized Benford's law in the wavelet-based contourlet transform (WBCT) is proposed in the paper. In the past research, the majority of algorithms frequently used a pre-determined threshold to tolerate the errors caused by some non-malicious manipulations. In contrast to the conventional methods, we use the adaptive thresholds instead of a fixed one for improving the accuracy of the tampering region detection. The adaptive process is based on the relationship between the compression ratio (CR) of JPEG2000 and threshold. Using the generalized Benford's law, once the CR is determined, the threshold can be adapted. The tampering detection rate is used to evaluate the performance of our scheme. Simulation results showed that our improved method can localize the malicious tampered region with more high accuracy.