Multi image codec filter - I'm stuck

Hmm, I'm not sure if its possible to combine more than two sources in floating point format (i.e. in integer and fractional parts) - I tried, but the "middle" part always get scrambled. I could be wrong though...

But even just with two sources, you'd need an external encoder if you want to import multiple images.

The "one image only" limitation is IMO the most annoying aspect of FF

Hopefully when we get bitmap scripts we can write our own loaders

Quote
I'm pretty sure that more than one image can be encoded. FForge definitely has the required bit depth (32/64 bits per channel?).

Yes, as I said, combining two images is no problem, but strange things happen when you repeat the process. I think it is because of the way floating point numbers are constructed (the values are not precise as with integers, they are approximations).

Sweet mother...
/facepalm
...it's soo simple...
/hides under the table in shame.

The original intention behind this was to batch together multiple
images for 3d work: diffuse, normal and specular maps.

Then use that composite image the usual way: paint it on 3d models,
transform them in PS, use the clone brush and so on. The final step
would be the extraction of the separate surface properties.

However I ran into a problem: Photoshop's "exposure" adjustment
layer has limited range so it has difficulties handling a dynamic
range this big. The result is an almost full white image which
is not very user friendly to work with.

I have a few ideas to try (storing extra data in the negative domain,
storing channels in RRRGGGBBB order instead of RGBRGBRGB, etc) but
haven't had the time to play with them yet.

I really wish we could lose the shackles of the 4 channels per pixel
limitation and have an arbitrary number of pixel properties...

The "negative domain" is really just indicated by one BIT in the floating point representation: either its negative or positive, it can't be both, so there is not like a whole extra set of bits that represents negative values.

However there might be a precision/range advantage in offsetting the final multi source composite so that it uses that BIT to double the precision. You can try it out by subtracting half of the max possible value (BITS^3 / 2 I think). Before you decompose the sources again in FF, you add that value.

HA! I just spent about 6 hours building a filter that does exactly this

Quote
Zoltan Erdokovy wrote: The original intention behind this was to batch together multiple images for 3d work: diffuse, normal and specular maps.

It was rejected as 'too slow to render'... which I found really fishy since it took 20 minutes to render on my little Dell laptop sporting a blazing core duo mobility chip.

If you want it, I can email it to you.

Ian

yeah, that's true. Though keep in mind this filter just helps you with the mechanics of combining 4 images into one, plus, it had been released that this functionality will be available in V3.

The main down side to multiple images via workaround is that coming in each of the four sources is 1/4 the working resolution enlarged to full size.
eg. 4 bitmaps put into a 2000x2000 pixel original to feed this filter would output a 2000x2000 pixel image containing 4, 1000x1000 bitmaps all now 4x the resolution. As long as you don't mind shrinking the FF to be 1000x1000 again. An imperfect solution for sure. So, I advise doubling the resolution of your source. This helps with the end sharpness... at least in my experience.

Quick3DVisualizor.ffxml

Cheers, I'll take a look when I can.

I'll also catch up on info regarding FF v3.