[Bf-committers] Re: A contribution

Dan Brown bf-committers@blender.org
Mon, 05 Jul 2004 20:03:20 +0100


GSR - FR wrote:
> dan@harthinian.com (2004-07-03 at 0018.47 +0100):
> 
>>.zpic - Output by softImage from what I can tell and importable into
>>AfterEffects, Combustion, and Photoshop. This is just a headerless
>>file of floats, 1 per pixel. I didn't find any more info than that
>>such as the range used and orientation but I will keep looking.
> 
> 
> All that other info must be derived from the PIC that has the colour
> data, so getting PIC info should help figuring orientation and such:
> http://www.softimage.com/community/xsi/discuss/Archives/sisdk.archive.0006/msg00004.htm

I've got a zpic file loaded and converted to a 16-bit raw file to test with. It 
is easy enough to do that. The main bit I can't test as I only have 1 test file 
is the range. The one I have (just a rendered cube) has values between 0.93 and 
100000 but I don't know if that is just for this render or if it is always the 
case. It may be that 1 unit represent one millimeter and the default camera was 
configured as 0-100 meters. That is the bit I need to find out, a relationship, 
if any, to the camera near/far clipping planes.

The resolution in the file is good, I am doing the equivelent of a contrast 
stretch in my code to make it fit the range 0-65535 (16-bit) ignoring the max 
depth as the cube itself is only between 0.93 and 3.5. I get a nice smooth depth 
buffer from it then with no visable banding.

I have also done a similar thing with the Blender depth buffer. That is stored 
'upside down' (first line is the bottom) and I get a range with a test pic of 
4billion to 4.1billion and with a contrast stretch get a nice depth buffer. I 
can play around with this one though to see the exact impact of changing the 
camera near/far settings, although I haven't yet.

Dan