To test some of our code (e.g. painting on canvas or importing/exporting different file formats), we need to compare the image data created by the operation under test to a premade image. However, our current implementation of the image comparison utilities does not provide stable results. The tests built upon these utilities are flaky and require a human to interpret the results. As a result, they these tests are run sparingly and we cannot use them for automated regression testing.

The goal of this task is to design and implement a solution to this issue, so that we can run such tests automatically on the CI.