The uniaxial absolute stress in steel members is a vital indicator for evaluating the performance of steel structures. Non-destructive ultrasonic detection is widely applied in the absolute stress measurement of steel members with the advantages of easy implementation, no limits on materials and high precision. The acquisition algorithm of time-of-flight (TOF), the sampling rate of ultrasonic signal and the processing of outliers determine the accuracy of the stress measurement. In this study, we proposed a norm and interpolation based algorithm for absolute stress measurement using the longitudinal critically refracted (Lcr) waves, which improves the accuracy and velocity of the absolute stress measurement. Firstly, the norm was designed to obtain the TOF of the ultrasonic signal. When only the Lcr wave arose in the ultrasonic signal, the two-norm similarity algorithm was found to be consistent with the cross-correlation-based algorithm and the corresponding theoretical expressions were derived. With the non-Lcr wave in the ultrasonic echo, the two-norm similarity algorithm has stronger anti-interference ability. Secondly, the pchipslopes, spline and linear interpolation algorithms were introduced via which the TOF of ultrasound with low sampling rate reaches the same accuracy as that of high sampling rate signal. Finally, a special class of outliers in TOF was explored, and a threshold based on the number of shift points and the stress feature was proposed to calibrate these outliers to normal values. Experiments demonstrate the proposed algorithm improves the accuracy and velocity of uniaxial absolute stress measurement in structural steel members.