normal mapped models

[rouncer]

im trying to do the technique where you take a high poly model and a low
poly model and project the highpoly models normals on the low poly model.

ive got it to work, but im not getting a very good result.

ive tried raycasting from the lowpoly surface to the high poly and ive tried
recording the highpoly models normals on a subdivided uv map of the low
polys.

neither worked that well, and the raycasting version seemed to do poorer
than the other method... and thats the way ive heard everyone does it.

what could i be doing wrong?

ive got a post at devimg.net if you want to see a picture of what ive got.

http://devimg.net/?Post=508

any help at all appreciated...

[Wernaeh]

It's kinda hard to tell what's going wrong by just looking at the pictures.

Some general ideas:

- The rough features seem to make the jump from high-poly to low-poly quite alright, so the base algorithm seems to work. Notice how the skull application on the helmet is off quite a bit in the right picture. Might just be a camera thing, but looks fishy to me. Some factor a bit off in one of your equations ?

- What resolution do you use for your normal maps ? Do you have mipmapping / filtering / etc enabled ? Are you sure the actual rendering of normal maps works the way it should ? Some of your highlights seem a bit off, as well, for example the bright white parts inside the upper edges of his armor. This sounds somewhat possible since you reported problems with two different algorithms.

- How do you calculate your UV maps ? Is there a unique mapping from space positions to model texels ? (Though considering the image, I'd say this is working the way it should...)

Guess there's not much to say apart from that, without actually analyzing the code you are using.