The quantitative assessment of software reliability by using failure time data observed during software testing in the software development is addressed. On the basis of two models described by nonhomogeneous Poisson processes, exponential and delayed S-shaped software reliability growth models, the authors adopt the mean time between software failures as a reliability assessment measure and propose a method of software reliability assessment. Since the distributions of the time intervals between software failures for the two models are improper, the mean time intervals are obtained by standardization of the distributions. Further, applying the two models to actual data, numerical examples of the mean time between software failures are shown. >