DNG noise reduction?

[AlefSin]

Hi,

I know, I know, many people have asked for noise reduction in RAW files before. I think sooner or later you'll need to do it as it is a real concern and has lots of potentials for you, as the developer.
Consider this: What is ISO noise in a CCD? Is the noise in adjacent channels in the RAW file (R, G, B or other colors) really that correlated?
When the bayer interpolation is done, basically the noise is now "spread" to the other channels and I can only assume that dealing with it gets more difficult. PS CS2 has good support for import/exporting DNG files. Now, if you could open, process and save back a DNG file in NeatImage, WIHTOUT doing the beyer filter (that is, working on the RAW file's channels separately) I think you could potentially:

• Make the noise reduction independant of the RAW processing method

• Reduce the niose more effectivley

• Make the noise profiles independant of the changes in WB or exposer or other settings in the RAW converter

• Make somthing unique, with no competition for the moment

Granted, not all DNG files are the same and possibly you won't be able to support ALL camera RAW files even if converted to DNG by photoshop. However, think of the large number of user only using Canon and Nikon DSLRs. For C1, it was big enough market to make a special version of C1 only for digital rebel; isn't it a big enough market for NeatImage?

Regards
AlefSin[/list]

[andewid]

Noise in a CCD/CMOS sensor is electrical noise, just like you get noise in other electrical equipment (your stereo for example).

The ISO setting would be about the equivalent to trying to play a very faint signal through the speakers by turning up the volume. You get more noise.

The noise from the sensor is also affected by the temperature of the sensor. The longer exposure time you use the hoter the pixels get and you get more noise. Eventually they get to hot and release false data. Then you get "hot" pixels, often seen against dark backgrounds in night shots.

I think the idea to use the raw invidual channels (pre-bayer interpolation) to do noise reduction is good. There are some problems though. You would not as easilly be able to separate colour information from luminance information as each channel would be equivalent to a grayscale image.

[NITeam]

Thank you guys. You are right, a lot could be done in RAW and I wish we could cover everything you suggest. In any case, I will take your ideas to our team meeting and hopefully something new will come out in the end. Thank you for helping us!

Vlad