4SAMPLES 3x3 5x5 PRE3x3 PRE5x5 SO3x3 SO5x5 - Explain?

Snex

Hey, just doing my normal map and instead of using crazybump I decided to use xNormal's photoshop plugin. Came across all these "Methods" and was just wondering if anyone could explain them to me, and which method is useful to certain types of models (hard edge, organic, etc).

The methods are: 4SAMPLES, 3x3, 5x5, PRE3x3, PRE5x5, SO3x3, SO5x5.

I did a search on google and polycount and couldn't find much.

Cheers

SpeCter

Pre = Prewit
So = Sobel

Both are Edge Detection Algorithms.
I´m not sure about the first ones, but i think they used a more direct approach calculating the normalmap.

3x3 and 5x5 would then pick a pixel and you use the neighbouring pixels to determine the height difference between them which gives you your z axis vector.

Unfortunately I have no idea what the 4Samples is.

If what i´m said about 3x3,5x5 then they are better suited for grainy details, because Prewit and Sobel could blur small details out.Most of the time Prewit/Sobel give more "smooth" results.

fearian

thanks speCter, and Snex, I have always wondered what these where about. Anyone else know more?

vvvv the latter :P

SpeCter

Do you want to have more in depth knowlegde or just more precise answers to what the different methods are?

Snex

anything would be good, I've also noticed that with the xNormal PS plugin, no matter what method you choose the normal map comes out really flat and noisy. Any way to fix this?

SpeCter

The outcome depends mostly on the source image. Diffuse images are not really suited to generate normal maps because there is lightning information in them and gradients you don´t want to have.A clean heightmap is best suited for that matter.

Do you have an example? Like the picture you are using and the normalmal it spits out?