Add an option to have approximate Gaussian points instead of solid points with a blur. #578
Conversation
Current coverage is 72.86%@@ master #578 diff @@
==========================================
Files 82 82
Lines 7131 7148 +17
Methods 0 0
Messages 0 0
Branches 0 0
==========================================
+ Hits 5191 5208 +17
Misses 1940 1940
Partials 0 0
|
…ints with a blur.
| * the python expression: | ||
| * from scipy.stats import norm | ||
| * for r in [0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1]: | ||
| * opacity = norm.pdf(r, scale=0.3) / norm.pdf(0, scale=0.3) |
There was a problem hiding this comment.
@manthey why apply scale to it?
There was a problem hiding this comment.
@aashish24, so that it has a sensible radius and fall-off. By using a scale of 0.3, the intensity at radius=1 is approximately 1/256th of the central intensity. This means that all of the pixels in the rendered circle have some impact on the display. If we use a a larger number (such as the default), the circle will still have significant magnitude at the edges of our rendered area. If we use a smaller number, then we won't use the whole area of our circle.
Simplistically, instead of using a scale, we could have scaled the radius. The mathematics are equivalent (we could have used norm.pdf(r/0.3) / norm.pdf(0)).
There was a problem hiding this comment.
Thanks for the detailed explanation and comment.
|
Lgtm (hit the wrong button on my iPad pro , sorry) |
This currently depends on the changes in PR #577. It could be separated if necessary.