Abstract: In ultrasonic reflection mode computerized tomography (URCT) a cross-sectional image of an object is produced by a reconstruction procedure from the envelope of the pulse-echo data collected surrounding the object. Thus, the echo of adjacent flaws in the object being detected will be confused with each other because of the existance of the transmitting pulse width, and it becomes difficult to distinguish the adjacent flaws in the reconstructed image. However, this problem can be well tackled if the time-domain deconvolution (TDD) algorithm is employed to process the echo data prior to reconstruction, and higher resolution for reconstructed images can be expected, as proved by our theoretical and experimental results.