This is a consolidating set of changes. It includes Ward's change. (merged) It adds unit tests and cleans up things a bit. The unit tests all pass. It also includes a set of test images as .png files. Manual system tests with test images show good results for 100x100px images and 120x120px. (50% as expected) But then 140x140px and bigger still gets odd result 48.6%. While not finished, I think these changes represent progress, and more testability.
It appears the accuracy issue with bigger images has something to do with the display or representation .
The two different algorithms yield the same results for the simple test images.
This is a consolidating set of changes. It includes Ward's change. (merged) It adds unit tests and cleans up things a bit. The unit tests all pass. It also includes a set of test images as .png files. Manual system tests with test images show good results for 100x100px images and 120x120px. (50% as expected) But then 140x140px and bigger still gets odd result 48.6%. While not finished, I think these changes represent progress, and more testability.
It appears the accuracy issue with bigger images has something to do with the display or representation . The two different algorithms yield the same results for the simple test images.