-
the maths to compare box blur + iterations, to fast blur?
so, i’ve been reading a few websites, including this one, https://prolost.blogspot.com/ and they seem to say that box blur with a few iterations will give a better quality blur for fine details than a fast blur (which is essentially the same as a gaussian blur, they tell me).
would any of you science guys happen to know how one could compare the two? ie, a fast blur of 5 px is NOT the same as a box blur of 5 px with 2 iterations, and that’s different to 3 iterations.
basically, if i want to use box blur, i’m guessing more than i used to when i used fast blur, because i can’t yet predict results, as, as i said, i was interested in the maths behind it. can someone fill me in?
-
5 Replies
-
Box Blur will do exactly what it says – it will blur pixel separately for each iteration. So if you have 3 iterations and a Blur width of 3 pixels, the first iteration will blur each pixel across an area of 3×3 pixels with centered weighting. After that it wil use the already blurred/ interpolated data and apply the other iterations. If you draw that on paper as a pyramid like that
your shape will look like a Gaussian curve (each o represents a pixel). The more iterations, the wider this gets.
Fast Blur basically is just one iteration. It will blur your image more because it only increase the size of the blur area, making the result less precise the larger it becomes (because the average value of all pixels becomes more “samey”). You loose a lot of detail and the blur increases linearly wheres Box Blur by its very nature acts more like a camera based Blur and maintains much more detail and crispness even at high settings.
Mylenium
[Pour Myl
-
Box Blur will do exactly what it says – it will blur pixel separately for each iteration. So if you have 3 iterations and a Blur width of 3 pixels, the first iteration will blur each pixel across an area of 3×3 pixels with centered weighting. After that it wil use the already blurred/ interpolated data and apply the other iterations. If you draw that on paper as a pyramid like that
your shape will look like a Gaussian curve (each o represents a pixel). The more iterations, the wider this gets.
Fast Blur basically is just one iteration. It will blur your image more because it only increase the size of the blur area, making the result less precise the larger it becomes (because the average value of all pixels becomes more “samey”). You loose a lot of detail and the blur increases linearly wheres Box Blur by its very nature acts more like a camera based Blur and maintains much more detail and crispness even at high settings.
Mylenium
[Pour Myl
-
Thanks for the explanation, Mylenium! So if I understand correctly, if I applied a 3pixel fast blur 3 times, would this be the same as a 3 pixel box blur with an iteration of 3?
Why don’t they have an iterations slider for fast blur then?Hmm perhaps I am perhaps still slightly confused.
Regards,
Scott. -
In principle you are right. The result should be about the same. I believe, though, that Fast Blur would still look different since it seems to mess with the sub-pixel smoothing (which Box Blur doesn’t – it works strictly linear). I don’t no and Stu doesn’t seem to know that for sure either, so I can’t provide further explanations.
Mylenium
[Pour Myl
-
Okay, thanks so much for the explanation. 🙂
Reply to this Discussion! Login or Sign Up