Fix corners

[scott-trinkle]

Addresses issue #5.

Slicing image.shape with [0:1] only selects the first dimension. So all four elements of the corners array were always scaled by the first dimension of the image, resulting in a potentially incorrect value for maxRadius depending on the center, which also leads to different image sizes for the polar image. Slicing with [0:2] uses both image dimensions and creates accurate corners values and fixes these issues.

I created a test, TestPolarConversion.test_final_radius(), that fails with the old slicing and succeeds with the new. With this particular image, the center used in test_default ([401, 365]) leads to the same value independent of slicing, so I used an arbitrary center of [350, 365] instead.

I noticed all of the tests were failing because the "horizontalLines.png" and "checkerboard.png" images are not included in the repository, so I commented those out.

Also, the indentation of the main function call in "test_polarTransform.py" was indented under theTestPointConversion class, so I don't think all of the tests were running. I changed that as well.

Let me know if you see any issues here!

[addisonElliott]

LGTM 👍

I am going to enable Travis CI and try to fix the tests issue so it's hopefully easier to test changes.

I made this library in like a week for a project. As more and more people find use in it, my motivation to maintain it will increase :)

Eventually I would like to make the tests a bit better and more comprehensive.