Fraction by matrix

Fraction by matrix
http://www.filterforge.com/filters/5424.html

Quote
ThreeDee wrote: Matrix, I don't know if I dare mention this, but there is actually a simpler way to do this:

Ah, that is indeed a much simpler solution...

I was actually going to update it with a third solution, but I think yours is even better:

Fraction 3.ffxml

Quote
matrix wrote: I think the bias curve may still be useful though, if you want to actually find the reciprocal value of an integer in the range 1..100. Maybe I should post that as a separate snippet...

Yes, there are definitely uses for the bias solution. You can actually use it to find reciprocal values in a range wider than 1-100, if you set the values differently. You'll have to move the decimal point before and after using the bias curve, but you can (fairly) easily extend it to calculate between, say, 0.01 and 10000. And there is a way to calculate the bias value. I recently figured it out graphically but I think it required integral calculus which I haven't done since high school, so I did the same as you and got the value by trial and error. I believe it was the same exact value that you got, too.

Ok I was really bored and simplified ThreeDee's Fraction 2 a bit more. It renders a tad faster now.

fraction U.ffxml

One down.

fraction V.ffxml

I wonder what is best in a performance perspective: using a tone curve and a linear curve or using the levels component. The levels component also needs to evaluate the gamma parameter, which might be expensive.

Quote
matrix wrote: I wonder what is best in a performance perspective: using a tone curve and a linear curve or using the levels component. The levels component also needs to evaluate the gamma parameter, which might be expensive.

Well I tested the render times and the levels method is a bit faster. I can't tell you what's going on internally though.