Piping classification to metamorphic testing: An empirical study towards better effectiveness for the identification of failures in mesh simplification programs

Mesh simplification is a mainstream technique to render graphics responsively in modern graphical software. However, the graphical nature of the output poses a test oracle problem in testing. Previous work uses pattern classification to identify failures. Although such an approach may be promising, it may conservatively mark the test result of a failure-causing test case as passed. This paper proposes a methodology that pipes the test cases marked as passed by the pattern classification component to a metamorphic testing component to look for missed failures. The empirical study uses three simple and general metamorphic relations as subjects, and the experimental results show a 10 percent improvement of effectiveness in the identification of failures.