Approach number Two
Look at center pixel's original weight
If sample'd pixel's weight is higher and within blur distance, blur
Otherwise, leave the pixel be
Idea being to allow the foreground to blur over the background
Weight Comparison Artifact
Problems
Sharp edge on ground still exists
Blur happens smoothly until edge is approached
Edge Blur Issue
When blurring near edge, center pixel lies on background
As blur occurs, background pixels are sharper, and don't get counted
One background pixel vs 75+ foreground pixels
Background overpowered, foreground seems solid
Approach Number 3
Look at the line of pixels being sampled, and their weight
If they blur far enough to overlay, count as usual
If not, recount the same pixel you are at
Idea being to offset the foreground overpowering the background
Considerably Better
Background still "crisp", even through blurred foreground
Foreground depths blur more seamlessly
A few Issues remain
Both caused by integer rounding, in blur difference calculation
Using floating point proved too slow to be practical