| raw image stacking before demosaicing | |
|---|---|
| algorithmically impossible | |
| algorithmically possible | |
| Login to vote and view results. |
0.00
#...
·
 |
|---|
|
Hi all I have really tough question: - do you know any techniques / software to stack multiple raw files before demosaicing them - and then export such stacked image as a native raw? [e.g. CR2 for Canon] |
10.79
#...
·
|
|---|
|
Hi, It's difficult to give a serious answer to your question without knowing a bit of more context. I'd assume you want to have files registered before? If not registered, you could easily do it in PI. You'd just have to turn off debayer for Raw formats and integrate straight forward. If registered, I don't see any serious way without debayering as you want to avoid mixing color channels. Björn |
|
0.00
#...
·
|
|---|
Björn: Well I am thinking about registered raws - that is the point and it seems very hard if not impossible to do. One may ask "but what for?" The answer is simple - for instance I am at night in the field taking astro-photos and I want to make handheld photo of my rig or my surroundings - I shot e.g. 16 raws - register them before demosaicing, export to native raw - and finally denoise them in DxO Deep Prime. |
|
10.79
#...
·
|
|---|
|
From a mathematical point of view, I would conjecture that it's possible but extremely complicated. However, I doubt that it's necessary. What would be wrong with demosaicing, registering, integration with outlier rejection, and subsequent post-processing, including denoise? |
|
0.00
#...
·
|
|---|
| Unfortunately DxO Deep Prime can process only native raws, no dng, no demosaiced stuff. |
|
10.79
#...
·
|
|---|
|
I don’t know this tool but I assume it simply wants mono data? Could you simply split the channels and process them independently? EDIT: I searched the web for it. Seems that it really only imports camera raws? |
|
0.00
#...
·
|
|---|
| I cannot split the channels. Only mosaiced raw can be processed in DxO. That is the point. I will be testing PI soon. |
7.90
#...
·
|
|---|
| Of course it can be done. Whether registering can be done meaningfully, that is another issue. |
0.00
#...
·
|
|---|
|
I know I'm quite late in this discussion, but... This is something I've been curious about for quite a while, my main motivation being that stacking channel by channel you eventually get "better" data (i.e. with higher SNR) to be demosaiced and I guess this could improve the performance of whichever algorithm you use. I see no conceptual limitation preventing you from trying this algorithm. I also have some ideas on how to carry this out with some well-known software. I will let you know if I manage to obtain something good.  |
4.06
#...
·
|
|---|
|
This is one of those times when something may seem like a good idea on paper but not really in practice. If you take a photo on a camera with a CFA the raw output is of course a mosaic which has yet to be debayered. If you want to stack that data it needs to be first aligned. That alignment process involves some interpolation. The result of interpolating a mosaic is going to be a mess because every 2 pixels is green and every 4 is red or blue. Therefore the data of neighbouring pixels is not even the same type of data. If the goal is to effectively de-noise, a far better solution is to stack the images after debayering. Then using the excellent denoise capabilities of modern processing software such as what is offered by Adobe, Topaz and even RC Astro. |
