Abstract: In the ultrasonic C-scan imaging, details are sometimes not clear enough due to the interference by external factors, such as noises. Thus, a method using AR (auto-regressive) spectral extrapolation was proposed to improve the quality of C-scan imaging. Compared with the conventions methods, the new method can increase the signal-to-noise ratio (SNR) and time resolution of the ultrasonic signals. First, the frequency window width was set by the AR spectrum extrapolation method. Then, a spectrum as the initial window was intercepted. Using the Burg algorithm, the AR model parameters were calculated, and the AR model was extrapolated to obtain the recursive spectrum. Finally, the inverse Fourier transform was used to obtain the time domain waveform. Taking the simulation signal as an example, the influence of the selection of the frequency window width and the starting frequency on the signal processing results was discussed. The analysis results show that the AR spectrum extrapolation method significantly improves the SNR and time domain resolution of the ultrasonic signal. The degree of recognition, the average result in the range of 3-10 dB frequency window was better than the processing results of other fixed frequency windows. This method was used to process the experimental data of a coin surface and a electronic chip. The C-scan images were compared to those by different processing methods. Results show that the C-scan image obtained by the AR spectrum extrapolation method is much clearer, the contrast is higher, and the boundary is more distinct. Furthermore, the imaging process takes less time. The results of the study effectively improve the imaging quality and efficiency of ultrasound C-scan.