CAMERA PRE-PROCESSING ORDER - POST SENSOR In-camera PROCESS

[superkraft]

I post this into the Post-processing forum, as there is no real pre-preprocessing forum:
I am wondering which are the processes done by the camera's processor for creating an 8bit video file:

As far as I understand it now, the following steps have to be done(or many more?), but which is the correct order?


CMOS Chip - Light detection: The matrix of G-R-G-B is read via a bayer sensor:
Camera Processing: 
- De-bayering 
- Converting 14bit raw to 8bit (how is this done in detail?, is it only clipping  or something like tone-mapping?)
- transferring image data the linear light to log (gamma)
- Adding the canon pictures style - presets (after or before the conversion to 8bit )
- Image optimizations (which ones do exist) sharpening process, ...
- Data transform to YcbCr
- Chroma subsampling process (4:4:4 to 4:2:0)
- Transferring the series of images into a movie container: Quicktime
- movie compressing: inter- and/or  intra-frame (mpg4, h264)

At which stage are the canon picture styles embedded?
Can I influence with the canon picture styles the conversion from 14bit raw to 8bit?

Thanks for any suggestions or ideas !

[superkraft]

Stevefal started a similar thread regarding questions of the "Canon image dataflow":

http://www.magiclantern.fm/forum/index.php?topic=4830.msg132823#msg132823

[chmee]

months ago i did some thoughts as well on this, maybe that can help:


regards chmee

[superkraft]

@chmee : Thanks for this interesting image !!!

Do I understand that right?: The Canon picture styles are embedded in the process to create the Low Dynamic Pictures, therefore the picture styles could change how the white and black levels are clipped ?

[superkraft]

Here another interesting image from Wikipedia on how the image-processing-pipeline of Nikon works.
I guess the structure of processing in Expeed will be somehow similar to the Digic process:

en.wikipedia.org/wiki/Expeed#mediaviewer/File:EXPEED_Milbeaut_M42.gif

[chmee]

@superkraft
as i mentioned, it was a first try. the nikon/fujitsu-chip is documented, unlike the digics in canon-bodies. to your question. i'm not that much into chip-specs and registers, but i think, black and whitelevel are needed as well for the a/d-path. so you can play with the canon picture style editor. this ones "should" give you the the answer, whats between 14bit sensordata and 8bit ldr. maybe theres another register or property we didnt find by now.

regards chmee

[superkraft]

thanks chmee !

Actually I think  your diagram is a  lot better than the Wikipedia one - it's clearer and simpler! ...  not even I understand the Wikipedia version in detail.
And maybe - if we keep this thread alive we can gather more and more information.