Title: Pseudo Distance Estimation Post by: JColyer on January 08, 2010, 05:50:36 PM I've been playing with Dave Makin's pseudo distance estimator (discussed here: http://www.fractalforums.com/3d-fractal-generation/true-3d-mandlebrot-type-fractal/msg7812/#msg7812) and noticed something odd. From the comments in the code it would seem that it's written to use a set of flat clipping planes. My code's based around using a clipping sphere and I noticed a lot of degradation (missing surface) when the rays are coming in from the oblique angles of the sphere.
I've been trying to figure out a good way to increase the accuracy of the estimation when coming in from the oblique angles for the past couple of days and I'm stumped. Any ideas? Thanks - JC Title: Re: Pseudo Distance Estimation Post by: David Makin on January 09, 2010, 02:25:59 AM Did you impliment the array check to avoid overstepping ?
You should note that using the "raw" figure from the pseudo DE, although accurate when very close to the surface, is not accurate enough for while approaching and I multiply the DE figure by 0.9 to give the step distance for quick renders and use a further reduction by a factor of at least 2 for final renders (in some cases up to 4 and occaissionally as much as 10). I do the same with the DE value when using the analytical method. This is a good view to test the speed/accuracy of your rendering code whichever DE method is used: http://www.fractalforums.com/gallery/?sa=view;id=1289 (http://www.fractalforums.com/gallery/?sa=view;id=1289) Title: Re: Pseudo Distance Estimation Post by: David Makin on January 09, 2010, 02:49:45 AM I've been playing with Dave Makin's pseudo distance estimator (discussed here: http://www.fractalforums.com/3d-fractal-generation/true-3d-mandlebrot-type-fractal/msg7812/#msg7812) and noticed something odd. From the comments in the code it would seem that it's written to use a set of flat clipping planes. My code's based around using a clipping sphere and I noticed a lot of degradation (missing surface) when the rays are coming in from the oblique angles of the sphere. I've been trying to figure out a good way to increase the accuracy of the estimation when coming in from the oblique angles for the past couple of days and I'm stumped. Any ideas? Thanks - JC Forgive the perhaps slightly patronising question, but did you normalise the direction vectors of the viewing rays so the steps are the same distance no matter the angle ? Title: Re: Pseudo Distance Estimation Post by: JColyer on January 09, 2010, 03:30:09 PM Thanks for the reply's David. I didn't include the arrays as I didn't quite catch their purpose, but I then saw another posting of yours where the array portion of the code was re-published and it started making some sense. I plan on implementing this over the weekend. And you're not being patronizing - i have no idea what I'm doing with this ray tracing thing! I'm just taking bits and pieces from here and there and trying to make some pretty pictures. But I digress, yes my direction vector is normalized when it's calculated from the x,y screen position. |