Dieter wrote:
It also points out some of what is wrong with the digital camera market:
1. Most cameras have 2 file formats JPEG & RAW. The first is suitable
only for snapshots and the second is really only for professional
photographers. Users that are serious photographers have to use RAW
since it is the only way to obtain a lossless image.
Apparently, even the "RAW" mode is heavily processed on most cameras. :-(
"The supposedly raw images produced by most Bayer-sensor cameras are in
fact heavily edited images prior to downloading them to the computer."
http://www.anandtech.com/digitalcameras/showdoc.aspx?i=3363
I think there should be a "TRUELY RAW" mode available. Provide the bits
straight off the sensor. (documented format lossless compression is ok)
The good news is that the Foveon sensor isn't as bad relative to Bayer
as previously thought. Now we need truely raw images from each to compare.
If JPEG2000 is used, there is no need for this two format mess. Since
the CFA image is actually monochrome, a lossless compressed JPEG2000
"RAW" image would probably be considerably smaller. If all the
meta-data is needed, the DNG format allows a choice of compression
method. JPEG2000 will also produce highly compressed lossy images that
look better for a given file size than JPEG since they do not have the
block artifacts.
You should be able to select what processing is applied to the image in
the camera. True RAW would be what you'd get if you select lossless CFA
with no other processing. Actually, you might want a True RAW image to
have defective pixels fixed since the camera knows this.
Noise reduction in the camera is not a good idea unless you have a
killer DSP chip in the camera and the time to do it. I can understand
why this is done to produce JPEG files, but I can not understand why it
is used for RAW on a professional camera.
The best way to eliminate noise is with homomorphic deconvolution. This
is a standard DSP process that requires a not of computations. With
radio transmissions, it is assumed that the noise is a stationary
stochastic process. The noise from a silicon sensor is not stochastic
although it should be stationary. So, you need to adjust the process to
use so-called colored (i.e. not White noise which is stochastic) noise.
This might take a while even on a multi-GHZ multi-core 64 bit personal
computer. This is an important point. At home you can have your PC
spend several minutes on noise reduction if necessary. You can't do
this in the camera so the camera's noise reduction will always be
inferior to what you can do on your PC.
Note, if the camera is going to reduce noise, the best method for a
camera is to use a Peltier cooler for the chip. The colder you get a
semiconductor, the less noise. This is why astronomers use liquid
Nitrogen cooled CCDs. Yes, you would probably need a battery pack and
it would take a while for startup, but this should be no problem for a
professional camera -- you could use if for standard daylight shooting
without using the cooler.
--
JRT
_______________________________________________
Open-hardware mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-hardware
