A generalized framework for systematic evaluation of algorithms.
A new image processing algorithm must be evaluated to determine its performance under different conditions. This is done by running the algorithm with different parameters on a provided data set. The output is then compared to the original or a ground truth image using a collection of metrics which are designed to probe the effect of the algorithm. The results are then stored in tables that tells conditions and results or the test. A final step is to compile the statistics and visualize the results.
The test procedure can also be used for parameter scans to identify the sensitivity to different parameters.
The documentation is found here