Software quality metrics of the Faint Object Camera (FOC) image processing software

The utility of software error counts and complexity measures in developing image processing software was assessed. The Rayleigh curve can be used as a predictor if the two parameters k, a, are known, but using a Rayleigh curve fit to data points at various stages of the Faint Object Camera image processing system project would have led to inaccurate predictions. Walston/Felix relations give simple global measures of a product which might prove useful as quality indicators. There is no evidence that complexity density can predict trouble spots in the code or detect problems associated with poor design decisions. The practical value of these metrics is limited and there appears to be no justification for stipulating maximum values of these metrics for a single module. Reliability models show promise for determining when a software system is sufficiently tested to allow release to the field. However, their use implies a test situation which emulates the situation in the field faithfully and may require a long (and expensive) test period. The models investigated appear low in predictive power.