Welcome to Fractal Forums

Fractal Software => Mandelbulb 3d => Topic started by: bib on February 23, 2011, 11:08:38 PM




Title: Color bands
Post by: bib on February 23, 2011, 11:08:38 PM
This color bands side effect is quite common when you use strong glow effects or when the 2 background gradient colors are very similar. It is not extremely visible in the attached picture, but sometimes it can be very annoying...  :'( :sad1:

Do you know how to avoid it? Is it due to a limitation of M3D? (24-bit colors?) :hmh:

Thanks  :embarrass: O0 :angel1:

Edit: this happens even when saving with 100% jpeg quality. Maybe I should save as bmp?


Title: Re: Color bands
Post by: Madman on February 23, 2011, 11:25:24 PM
I think it's a 24-bit colour limitation. I recognize the issue (mainly with gradients of similar background colours in M3D), but have also encountered it with  in UF and Chaospro if colours are close to each other. I think it's time for the introduction of HDR in fractal generation! And cheaper true HDR monitors...


Title: Re: Color bands
Post by: bib on February 23, 2011, 11:48:05 PM
I have also encountered this effect in other programs but I think that it happens more frequently in M3D, so I wonder if it can be enhanced or if it's just me.

Jesse, if you read me?...

Have a look at the other example in attachment (an extract from a bigger picture)...I feel that 24-bit colors could do a nice fade here, but that M3D does not do it.

Maybe I like this program so much that I ask for too much :D


Title: Re: Color bands
Post by: Jesse on February 24, 2011, 12:20:41 AM
PNG would be an option  :dink:

If you are talking about the "dynamic fog", that is based on the raymarch stepcount and so it is naturally integer based.  Here the "first step random" option can help if there are only few steps.

The background gradient should not be worse than any other common 24 bit color image, and the maximum the current displays can show...
but i will look in the painting routine if i find some issues (but the values are treated in floats in here and the last step is rounding).

More visible accuracy than 8 bit per color channel would be only with dithering possible.


Title: Re: Color bands
Post by: mondaymorninglunatic on March 10, 2011, 05:41:10 PM
yes! a dithering option would just be great!


Title: Re: Color bands
Post by: Jesse on March 10, 2011, 08:19:57 PM
Did you tried the "first step random" option, it is like a dithering for the dynamic fog banding issue.


Title: Re: Color bands
Post by: marius on March 10, 2011, 10:23:13 PM
Did you tried the "first step random" option, it is like a dithering for the dynamic fog banding issue.

I have been battling color banding in the distance w/ gpu shaders as well.

So, generally i think?, level of detail (and performance ;D) management is typically done by increasing minimum distance bailout of the iterative ray marching, depending on projected on-screen error.
That means the drawn 'hull' around the fractal is thicker in the distance. Hence coloring based on magnitude of final vector and such will 'band' since for say a smooth surface at an angle in the distance, since the computed distance to the surface for the pixels will vary and hence the color. Very distracting in movement / movies, dithering / aa or not.
Fog or other muting of color in the distance, or limiting depth of field (i.e. introduce blur) is a general approach to make the effect less pronounced, as well as to control high frequency aliasing noise.
Are we talking about the same problem? Can we do better? At speed?