48 Bit Dithering...

Dave Savage

48 bit dithering In render options...

What does it do?

Several renders I have done, I tried it and there was no real difference except it seemed to make the render time longer.
So was wondering; are there some render settings where it helps and are there some render settings where it speeds up the render time?

Thanks.

cris333

From what i read (can't remember where, i think on bryce guide) ,48 dithering blends the colors. It's similar with Blend with Sun option but it's working on the entire render image and blends colors when there is a difference between colors passes. So if we have a red cup on a black/blue table, 48 dithering will/should create a purple 1-byte (or more) line on the cup's edge surrounding it.
I think It makes the colors to blend on the edges by rendering on a 48-bit color by creating new bytes of colors ...? highest color resolution on web is 32bit right ?

I'm not sure , i didnt had time to test this closely , wait for a reply from a bryce master or anyone.
But one thing is for sure, as you said, rendering time is highly increased, this i noticed, without an important difference.

I think this can be checked with photoshop by magnifying the images(final renders) to compare them at ~3200%.(at byte level)

Edit: thank you elvis. also 48 dithering is good on heavy loaded scenes such as sci-fi cities . example here : http://wall.alphacoders.com/big.php?i=91852 .

Electro-Elvis

cris333 has described it quite right. I found following explanation (incl. the gamma correction) which Horo wrote in another forum, hope it helps

Gamma:
For the sake of simplicity, lets consider a monochrome (black/white) picture. Example: black has the value 0, dark grey 64, grey 128, light grey 192 and white 255. The difference of each step is 64, black brightens gradually to white along a linear (steady rising) ramp.
Now gamma makes this a logarithmic ramp. Black has still 0, dark grey has already 128, grey 192, light grey 224 and white again 255. The dark parts of the picture are brightened more than the bright parts and get more shades of grey in between. With an inverse log, you do the opposite: get more hues in the bright parts and less in the dark parts. This is all true for colours as well, of course, since colours are mixed. Each of the colours red, green and blue can have values between 0 (dark) and 255 (bright) and there is a linear or logarithmic function for each one.

Dithering:
Used to distribute image contents when a small picture is scaled up (or printed). There are many different dithering algorithms that give different results. Without dithering, the pixels would grow to boxes.
Imagine having a black and a white pixel adjacent and size it up by 2. Without dithering, you get 2 adjacent boxes 2 x 2 pixels black and white. With dithering, you would get a black (0), dark grey (85) a bright grey (170) and a white (255) one. If you size up 4 times, you'll get additional pixels in between.
What about sizing up by 1.414? You cannot work with integer numbers anymore but you have to resort to reals with good precision to calculate sub-pixels and sub-brightnesses. 48 bit in this context is an odd number, even if it is an even one but it shows all is done with appropriate precision.

David Brinnen

Long story short. If you are saving at 24 bit colour precision (the default .bmp option), then 48 bit dithering attempts to create the illusion of 48 bit colour by granular blending, much as you could create the illusion of grey from a distance with closely packed black and white dots.

If you save your renders using the export image options as either 48 bit tiff or hdri - then you don't want this option turned on as it only serves to add unnecessary "noise" to your renders.

Horo

Another quicky: Bryce renders using single precision reals. These use 16 bits to represent a value from about 3 x 10^-38 to 3 x 10^38. So there are 16 bits to represent each primary colour, which adds up to 48 bits. If exported when the render has finished, you can save the full range of 48 bits. Otherwise, they are reduced to 8 bits integer values. And here, dithering comes in.

There are indeed many different dithering patterns and Bryce uses one of them. Since a lot of fine details in a transition from one to another colour are lost when reducing to 8 bits, dithering attempts to hide part of the steps. Just like a printer. Look at a print out with a magnifying glass. Just as David said, use 48-bit dithering if you intend to save your renders as a conventional BMP. If you intend to export it as 48 or 96 bits image for later processing, do switch it off, because it introduces a pattern that renders the image useless.

We have to be careful to understand what is meant with those bits. A conventional BMP (or JPG, or PNG) has 24 bits, 8 for each primary colour. A 32-bit BMP has still only 8 bits per colour, plus a fourth 8 bit channel that holds the alpha map.

48 bit TIFF has 16 bits per primary colour (single precision real) and a 96 bits TIFF (or HDR) has 32 bits per colour (double precision reals) - and GeoTIFFs use quad precision reals (64 bits) to represent greyscales with a very fine resolution.

Dave Savage

Some of that makes sense... thanks all. :)

I mostly save my renders as .hdr and .tif but much like the "gamma correction" I've never really understood it's use (until now).

Horo

I'm not a fan of gamma correction, though it might have its uses. More info about it http://en.wikipedia.org/wiki/Gamma_correction. Bryce gamma correction has no user control and to my knowledge, it doesn't work exactly the same on the Mac and PC.