Sensor blooming revisited

[didger]

OK, I've cleared some space on my website and put up a high quality hi res jpeg showing the worst of this garbage.  If you think that this could conceivably be fixed, but maybe not provable with this jpeg, I could send you a CD with the raw file and a full size tiff, or if you have a location I could FTP the material to, that would work also.  In any case, maybe seeing the picture gives you some idea of exactly what's going on.

[didger]

This posted image is at least 1/2 stop overexposed if you look at the rocks in the background.

Yeah, but normally 1/2 stop overexposed is no big deal and conventional wisdom says shoot for the right end of the histogram to reduce shadow noise and I can't see just blue over on my 1ds field histograms either.  I do leave my camera permanently in saturated pixels flashing mode, so it's a little strange that I got neither any histogram or flashing blown pixels warning with this.  I've only seen this weird color noise problem twice now, both times shooting into the sun where there were very bright highlights.  Now I know that in the future I'll underexpose at least a stop and check to make sure that the histogram isn't then falling off the left side on those very rare occasions where conditions are such that this is liable to happen.  I'll also experiment with a polarizing filter.

I'm still a little confused about just what it is that you are seeing

Yeah, me too.  It just looks like a totally weird wrong colored unidentifiable layer of some sort of garbage.  In some of the other frames this time and on the other occasion it was more clear that it was very screwed up areas of densely clustered highlights on flowing water.  In any case the actual scene had nothing that looked like this whatever it is.  I tried all sorts of things removing these green and purple colors (successfully), but even converting the image to greyscale leaves that area looking weird, pixillated and unidentifiable as to what it might actually be.

It's definitely looking like this is a prevention and not a cure situation.  Something I had to go through in the process or racking up my 10,000 hours on the way to technical mastery.  :p

[Robin Balas]

Yeah, but normally 1/2 stop overexposed is no big deal and conventional wisdom says shoot for the right end of the histogram to reduce shadow noise and I can't see just blue over on my 1ds field histograms either.  I do leave my camera permanently in saturated pixels flashing mode, so it's a little strange that I got neither any histogram or flashing blown pixels warning with this.

The histogram on the 1Ds is "colorblind" it shows the green channel only. Hence if shooting bright blue or red you will blow those channels without warning in the info view. I recently documented a large stained-glass window in a new church and the blue was 5 stops brighter than the reds. The green channel was below that. My luminosity histogram and exposure went nuts, but when enabling the RGB histogram on the MarkII's it becomes a perfect aid to the correct exposure - although heavy post processing was needed to reproduce what I subjectivly saw. The reds and greens had to be brought several stops up from the shadows.
The contract proofs showed me that trying to reproduce the vivid glow of dense blue glass is almost impossible, and we had to settle for a rather boring interpretation of a great piece of art. The artist was very satisfied though and I couldn't really disagree with that. But it is very important to be aware of the primitive luminosity or green channel histograms limitations, at least when in high altitude or in extreme warm lighted rooms like theaters or night clubs.
MHO