I spent an interesting evening at Antwerp University last week, I’d gone to do a short presentation on HDR & WCG.
The main presenter was a scientist who was comparing raw with log and “baked in”, he went on to “Prove” that you can grade “baked in”just as well as you can raw, in fact it could be better he implied.
Well excuse me! that may be true in a theoretical world where we have infinite bit depth but in a real world situation that just isn’t true.
The limitations of the formats we use will cause damage to the image unless we are very careful.
I tried to point out that raw was always better in a real world because the camera output required processing and the processing we could do in camera was limited compared to what we could do in post. The cameras inbuilt processing was limited both by size and heat generation.
Apparently not from a scientific point of view, he could prove I was wrong, here was the proof on the whiteboard.
Well excuse me but I make pictures not numbers on a whiteboard and I can clearly see the difference. We briefly touched on compression and “lossless” compression, I gave up, the argument would have taken the whole evening.
2 stories for background to my point of view…
I was shooting test footage for Kodak at their plant in Chalon France, I had my entire commercials crew with me and we were looking at 2 prototype films compared to 5294 and Agfa XTR250. The reason we were including the Agfa was that it was starting to eat into their sales and they wanted an emulsion to beat it. I was a heavy user of XTR 250 as were the 11 other DP’s shooting tests worldwide.
One of the prototypes was just awful but one of them was looking good. I used the XTR because it had much greater latitude than the ’93 which was popular at the time, a lot more latitude and it fitted my lighting style.
We set up a final scene to stress the stocks which involved a model moving through light and dark in a Chateau and tested highlight colour separation.
Looking at the results projected it was very quickly clear that ’93 just couldn’t cope but that the good prototype had better highlight and shadow detail than the Agfa and more accurate reproduction of the very saturated red dress the model was wearing, it also coped really well with the subtle highlight colours.
After I’d said all that the Kodak guys said “so you’ll switch to this when we introduce it” my reply shocked them, “no, it’s not as pretty”.
My job is not to make accurate pictures it’s to make pretty ones…
Next story, years ago I was at NAB and it was all getting to me so I escaped to the sports bar in the Hilton next door (it was on you right as soon as you left the convention center) as usual it was full of other escapees. Also as usual I got chatting to a guy at the bar, I didn’t know who he was because his badge wasn’t showing, he was just Ray.
We had a great conversation about compression and he obviously knew what he was talking about. We talked about lossless and visually lossless. The first being what it advertised and the second depending on the judgement of the mathematicians who wrote the algorithms used.
I took away the clear point that anything over 3 to 1 was not true lossless and therefore depended on someones opinion.
As we were leaving his badge appeared, yes, his name was Ray, Ray DOLBY! I kinda thought he knew what he was talking about 🙂
Now, is someone who is a mathematician/scientist as aware of what makes a good picture as I am? why would anyone make that assumption? they don’t assume I know as much about maths as they do, how would they react if I came along and told them their algorithms were wrong and altered a few lines of code?
Treat me with the same respect I treat them, I accept they know their business better than I do, maybe they should accept that I know more about how to make a good image than they do…
Geoff, I have no idea who this ‘scientist’ is/was, but I would place his assertions under ‘bollox’.
Even in the theoretical world, the best images will always be those that have had the least processing applied.
To suggest otherwise means you are also bringing in an element of ‘personal’ input that below a given threshold the degradation will be below a visible threshold.
Who’s threshold?
The same is true of ‘lossless’ compression.
Unless the image after compression is binary identical to the original, it is not lossless.
At what level such losses become visible is a separate question.