Regression Testing: Test Cases for Graphical Images

The paper proposes regression testing solution for systems with several thousands of test cases to be run. The system generates patterns - many various images consisting of SVG objects - lines, circles, etc. The image diversity makes impossible to carry out testing in a traditional way - by recording test cases and then replaying them. The authors propose (1) For each graphical image (pattern), image-specific parameters shall be identified, allowing for regression testing to automatically identify significantly different images from the benchmark images. The value of the parameter that describes the images is calculated first, such as the overall length of the image lines, the area, and so on. A slight difference between the result and the calculated parameter is acceptable. Otherwise, the test has revealed a non-compliance to be assessed by experts. (2) Support for testing (instrumentation) is incorporated into the system, including code fragments that allow changes in SVG objects' selecting order. This allows to track the process of generating a graphic image and thus identify the causes of the difference from the benchmark images. The proposed approach is demonstrated with the help of a practical example. Other areas of application could be, for example, in cartography, where differences in maps are assessed, or in medicine, where the state of human health is assessed after visual changes in organs.