Canon Line Offset Noise

Started Oct 20, 2009 | Discussions thread
ohyva Veteran Member • Posts: 6,342
Re: The question I've got is...

So if I understood you right, you propose Canon should manipulate the sensor raw data before writing the RAW file to memory card. I'm definitely against that as there may be better ways to do the data clean-up than what would have been implemented in the FW. I do not really mind what the RAW files look as long as they can be converted the best possible images.

If the correction is that simple, every RAW tool should implement that with no difficulties. And Canon to do something similar in their FW for JPEGs. And the best tool developers can then implement perhaps even better ways to handle this.

John Sheehy wrote:

JimH wrote:

The question I've got is this:

Could a person come up with a good average value for these offsets for a given copy of the camera such that they'd be useful for other frames taken by the same camera?

Probably, I would say, for this particular pattern. A simple formula based on the modulo of the x value would probably work, with a factor for ISO if necessary. The approach I used would work for any whole-line offset, and need not know about ISO.

If so, and if things don't drift or change with temperature, then a person might well be able to feed a few black frames into a converter program for analysis, store that camera's "black profile", and then have it automatically massage our CR2 files from that camera before we even suck them into DPP or ACR.

Yep. Just like dust data, we could have gain and offset corrections.

This could be automated to operate on batches of files (say, everything in a directory). So it'd be an easy thing to apply to all of our shots to remove this problem.

Yes, but it would be nice if the camera calibrated itself, and this was all done before the RAW is written. 14 bits should be enough for doing the necessary corrections, but a mediocre 16-bit ADC with 13 bits of real precision (3 for dither - better to use noisy bits in an ADC than to "add" dithering noise) should be sufficient for the calculations, without causing histogram shifts.

As far as the other main issue, gain, is concerned, the 7D is already compressing the original RAW histogram, so just changing the factors to equalize the channels would add no further problems.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow