Title: Failed DE over 2900 % - what am I doing wrong Post by: darksky on September 20, 2011, 01:28:00 AM Here is my config: http://pastebin.com/drcrxdgU
I can't understand what I'm doing wrong. The above is my first attempt at a keyframe animation. It starts out okay, but as I zoom in, I get increasing values of Failed DE. I thought by optimizing the DE step factor, I would avoid this. Thanks in advance. Starting image: http://postimage.org/image/2fzblmc3o/ Middle image: http://postimage.org/image/2fzewp138/ Title: Re: Failed DE over 2900 % - what am I doing wrong Post by: taurus on September 21, 2011, 09:55:57 AM hi darksky and welcome to the forums. :howdy:
i think you did nothing wrong. while i don't know what DE error means, i see this value often in absurd heights (what the hell is 2000% ?). i ignore this, as i never encountered any quality loss. the important value in this row is missed DE. this is the value affected by the optimisation and it should allways be lower than 1% at least (better is 0,10%). have fun! taurus Title: Re: Failed DE over 2900 % - what am I doing wrong Post by: Buddhi on September 21, 2011, 06:23:19 PM hi darksky and welcome to the forums. :howdy: i think you did nothing wrong. while i don't know what DE error means, i see this value often in absurd heights (what the hell is 2000% ?). i ignore this, as i never encountered any quality loss. the important value in this row is missed DE. this is the value affected by the optimisation and it should allways be lower than 1% at least (better is 0,10%). have fun! taurus I have just realized that there is some small bug in the program. This value "DE error" in console window shouldn't be given in percent. This is total number of pixels where was overstepping problem caused by wrong distance estimation. The value "MissedDE" on progress bar is calculated in the same way (but properly), but this is percentage of pixels with overstepping problems. When there is high value of "MissedDE", there is a lot of noise on image (many black pixels, high roughness of surface or missed some parts of fractal). To reduce this problem sometimes is needed to decrease "DE step factor" (not too much because rendering will be very slow). Title: Re: Failed DE over 2900 % - what am I doing wrong Post by: darksky on September 22, 2011, 01:45:16 AM Thanks for the replies, all. I'm glad to try the new setting. As you can see in screenshot #2, there is a lot of noise in the images. I will try the setting you suggested!
Title: Re: Failed DE over 2900 % - what am I doing wrong Post by: darksky on September 23, 2011, 03:21:38 AM the important value in this row is missed DE. this is the value affected by the optimisation and it should allways be lower than 1% at least (better is 0,10%) If there is no mention of this parameter in the output should I assume that I am not missing any? Code: $ mandelbulber -keyframe -start 2 -end 3 -format jpg ~/.mandelbulber/settings/try1 Also, I still have black dots all over my image, even with a DE Step of 0.08 which is what the HQ optimizer suggested. http://postimage.org/image/fmxewrno/ Title: Re: Failed DE over 2900 % - what am I doing wrong Post by: taurus on September 23, 2011, 02:12:47 PM the "missed DE" value is located at the progress bar of the main window. "Failed DE = 32889%" should mean almost 33000 pixels failed for distance estimation (see buddhi below). depending on size of image this might be a few percent. the optim. buttons do not allways work propper, so i suggest to decrease DE step further. try for instance 0.02 and watch the progress bar...
you might also post the entire params here easily with the copy/paste button. good luck! taurus Title: Re: Failed DE over 2900 % - what am I doing wrong Post by: Buddhi on September 23, 2011, 05:22:04 PM the important value in this row is missed DE. this is the value affected by the optimisation and it should allways be lower than 1% at least (better is 0,10%) If there is no mention of this parameter in the output should I assume that I am not missing any? Also, I still have black dots all over my image, even with a DE Step of 0.08 which is what the HQ optimizer suggested. http://postimage.org/image/fmxewrno/ It's because this formula is generally difficult to render. There is no accurate distance estimation for this formula in Julia mode, so ray-tracing is not accurate. There is no way to make this better. |