Fix elliptical gaussianBlur

[gr5]

There's a feature where you check a box in "configuration" "mirror" called "elliptical flat or nulled optics".

When this is checked and you check gaussian blur in results tab, it turns the edge.

This can be fixed by changing from cv::gaussianBlur() to CropGaussianBlur() but you have to pass the vertical axis ratio to cropGaussianBlur and you have to modify CropGaussianBlur to take one more parameter and deal with making the elliptical mask.

If someone does this you need to pay extreme attention to the size/shape of the blur mask because if you are off by a pixel on the radii (vertical and horizontal) you will get bad results; results visible in the 3d view.

You might have to shrink the outer border by 1 or 2 pixels to match the data going into CropGaussianBlur.

Elliptical feature is not used much so this was put on the back burner.

[githubdoe]

Like George said. Flats are not measured using a Bath. That is important to remember because they do not create the background noise that a Bath does and then almost never need Gaussian Blur. Second very few users use that feature. Just leave it.

[githubdoe]

At one time Openfringe used int coordinates. That was a bad idea and caused many problems particularly when the mirror outline diameter was even. In that case there is no pixel at the center. This always caused a fringe at the edge of many operations.

The solution developed by Steve Koehler and I think Mike was to set the center of the coordinate system so that 0 is in the middle of a pixel and not at an edge as using int's do.

[gr5]

Dale I think the idea was not to change how the center and radius is stored but simply the one function that creates the mask - it could convert the numbers to integersX2, use an openCV function to create the mask and then discard those integers.

[atsju]

Wrong thread it's #275

[gr5]

lol, right. Well the 2 comments above by me and Dale should be in issue #275