[Leica] Digital: colour rendition and image quality

[jonathan at openhealth.org (Jonathan Borden)]

Peter Dzwig wrote:

> Dear All,
>
> I have just spent a few days photographing and recording some of my 
> late father's paintings before they were sent on semi-permanent loan 
> to the University of Torun in Poland.
>
> I decided for fairly obvious reasons to use my D-1 to take digital 
> images of the  70-odd paintings which have been shipped. I was 
> horrified at the variation in colour between the painting itself, what 
> I saw on the screen and what I actually got on the SD card. Of course 
> I am aware that my eye and the camera don't have the same response 
> characteristics; but interestingly I could do little to get the camera 
> to come close...
>
> Is this my D-1, is it common or is there some reasonable explanation?

Don't be at all surprised. This will *always* be the case whenever you 
attempt to accurately reproduce colors by digital or film or any other 
technique.

You must use a profiled workflow. What you can do is take a shot of an 
IT8 target with your camera. This will allow you to get an ICC profile 
for the particular lighting environment w.r.t your camera (you want to 
use RAW images).

You need to profile your monitor, printer, inks and paper.

When done you will get *more* accurate colors, assuming the gamut of 
your imaging system is adequate to capture the gamut of the paints.

For example, the gamut of Ektachrome is larger than the gamut of any 
printing technique -- particularly with bright yellows as I recall.


>
> Further, my father painted largely abstract works (you might describe 
> him as an abstract expressionist, but it's not particularly accurate). 
> On those where the boundaries between areas of colour were not 
> distinct the camera appeared to have difficulty in producing a sharp 
> image. The opposite being true where there were strong boundaries. I 
> am coming to the conclusion that the image-reconstruction algorithms 
> taking the output from the chip and building the resultant image in 
> memory must have been fooled. Any thoughts??
>

Perhaps the images are not in focus? Also this is where you really need 
hi-bit e.g. 16 bits/channel images. Very subtle changes in color can 
get quantized away with 8 bits/channel.

Accurate reproduction is harder than you might initially consider.

Jonathan