Aimed at ultrasonic signal compression and detection of the flaw echo of aluminum alloy forge in the presence of high scattering microstructure noise, the adaptive stein unbiased risk estimation (SURE) compression method based on the new thresholding function was presented, and the flaw model under ultrasonic signal has been built. The new estimated wavelet coefficients were deduced by using the new thresholding function, as the original wavelet coefficients decomposed from signal inputted. The new estimated signal was rebuilt by using the new estimated wavelet coefficients. An iterative process of the thresholds, which depend on the minimum mean square error (MMSE), was implemented to decide the termination of signal rebuilt. The experiment results indicated that the adaptive compression method has better de-noising and compressing performance and improved the SNR ability for flaw echo detection.