Something just aint looking right
The green path to the left shows results from the bright-pass and bloom filtering. The blue path on the right shows the luminance measurements. The red paths show the original raw HDR data and the final composition.
(Click To Enlarge)
This image shows the side of the cube thats filled with regular 0..1 range colours. Being all LDR there is no bright-pass/blooming, and the luminance part is fairly boring. Final image looks pretty much like you'd expect the "classic" spinning cube to look.
(Click To Enlarge)
This image shows the same cube rotated slightly so that a couple of the sides with HDR (~10.0 values) colouring. Because the exposure is still relatively low (note the still dark grey luminance results) there's a lot of bright white and a bit of bloom/blurring appearing on the edges.
(Click To Enlarge)
This image shows the cube rotated so that the 3 visible sides are all HDR-coloured. The important part is that the luminance measurement is substantially brighter, so the "true" colours of the HDR surfaces becomes more apparent. Not entirely sure what/where the bloom/blur went in this one [smile]
So, thoughts?? Anyone spot anything particularly wrong with those results?
All the code says it's but I'm not convinced [headshake]
- User interface needs working on - cells are too small, far too much whitespace. Might completely re-write that part.
- User interface needs a scalar for presentation - too much over exposed white in the intermediary cells
- Bloom effect is very limited in range despite operating on a 4x4 downsampled image. Can't seem to figure that one out [sad]
- I guess my hardware can't do linear filtering of an HDR texture, no matter what I try it will only stick to point filtering.
*beep beep ba beeeep ba ba click click shzick!*