[interested_observer]

Last week I went out shooting the Milky Way. I'm still processing the take (with a brand new different work flow, and utilities). I had processed these shots - actually panoramas and in looking at them I decided that they may be interesting to post. Both were taken with the K1, DFA 15-30/f2.8 @ f8 infinity focused, on a tripod from the same exact place - and 7 minutes apart (and someone turned off the lights during the break). They were both stitched and processed in LightRoom Classic.

The first one was bracketed @ +/-2ev

The second one was using Pixel Shift (which automagically DeMosaices the PS file and renders the image using all 4 PS images)

I've done a little work with Pixel Shift before, but in looking close I saw some remarkable detail that I had not seen earlier (well my old copy of LightRoom 5.6 didn't support PS, and also I just started using Raw Tharapee). So, I was going off the results in the in-camera processed image.

This cholla (and you absolutely do NOT want to even get close to them - as they will jump out and lodge their fish hook spines in you [and yes I had a pair of needle nose pliers in my backpack]) is about 10 feet away from the tripod, on the extreme left side of the panos. Yes, the lighting is dramatically different, but even in the reduced lighting environment the detail is substantially improved. Now, this was not intended to be a scientific comparison is any way - even with all the differences, I was just really impressed.

I'm going to pretty much bypass bracketing and just go with PS, unless there is some reason not to. Also, just to note that the dynamic lighting involved with the PS when I was shooting, there are some very fine artifacts due to the changing lighting conditions. The artifacts are so small, they are nearly impossible to see unless you go to about 300% looking for them.

• Pixel Shift with LightRoom Classic CC & Camera Raw - PentaxForums.com

______________________

Just for grins - I added two chipouts from the background. Due to the changing lighting conditions, there are some features that are better on one, while other features are better on the other....

Attached Images

       

---

Last edited by interested_observer; 07-11-2019 at 09:22 PM. Reason: added last 2 images

[jbinpg]

Nice stuff. Just a quibble: since PS files already contain all color information at each pixel, they do not need to be demosaiced. This is one of the reasons they lead to better files all around.

[biz-engineer]

The second one was using Pixel Shift

More details but also shows lines/patterns in the blue sky area of the image. I prefer the color palette of the first image.

[interested_observer]

Nice stuff. Just a quibble: since PS files already contain all color information at each pixel, they do not need to be demosaiced. This is one of the reasons they lead to better files all around.

Thanks for the comment, and I do understand what you mean. One of the largest problems that I've had in the last week after moving to the LR Clasic v6 (and Adobe Camera Raw) was with respect to how it was handling the PS files, since there is no indication of what it actually does (i.e., recognizing that it's a PS file), or like Raw Tharapee does by the user having to specific the type of demosaiced method to use in their pull down menu. Also, there is no information on the web on this at all.

The way I determined that LR6/ACR was processing the 4 component images was the small artifacts in the file due to the rapidly changing light which essentially creates a sort of motion effect all of its own. Also, the in-camera rendered PS image is a bit more flat than either the LR6 or RT processed image.

My problem was how do I describe the processing of the component images by LR6 - so I went with the demosaiced label to get the idea across. Yea, it's wrong - but the best I could come up with at the time.

[BigMackCam]

More details but also shows lines/patterns in the blue sky area of the image. I prefer the color palette of the first image.

Agreed. In fact, if you look closely, those lines and patterns are also visible in the land areas too.

Is this an artefact introduced by the processing software, or is it somehow inevitable with pixel shift images?

[Rondec]

@heckflosse may be able to give some input into the situation.

I will say that even with changing light conditions I don't see the artifacts you got. Typically in Raw Therapee I choose pixel shift and then take a look at the motion mask. If it looks like movement is more than about 50 percent of the image, I don't bother with it. I then choose whichever base image is the sharpest and in your situation I would click the "equalize brightness of frames" button. At that point I save the image as a TIFF file and go from there. There are some other things you can do with pixel shift out of auto mode in RT, but I don't use those much except occasionally median mode when shooting moving water.

I've found that pixel-shifting has improved my ability to calibrate lenses, too.

[interested_observer]

More details but also shows lines/patterns in the blue sky area of the image. I prefer the color palette of the first image.

Yes, they are pretty subtle though. I didn't realize that the lighting was going to change so quickly. When it turned it was red one second and blue the next.

Agreed. In fact, if you look closely, those lines and patterns are also visible in the land areas too.

Is this an artefact introduced by the processing software, or is it somehow inevitable with pixel shift images?

Yes, they are all across the image. I know that it's not motion and just the change in light, when I saw that they were across the rock faces on the mountain. It's pure lighting. I found this link that is probably the best explanation that I have run across yet.

• What happens if you try to pixel shift a long exposure with movement. ? David Abbs Photography

@heckflosse may be able to give some input into the situation.

I will say that even with changing light conditions I don't see the artifacts you got. Typically in Raw Therapee I choose pixel shift and then take a look at the motion mask. If it looks like movement is more than about 50 percent of the image, I don't bother with it. I then choose whichever base image is the sharpest and in your situation I would click the "equalize brightness of frames" button. At that point I save the image as a TIFF file and go from there. There are some other things you can do with pixel shift out of auto mode in RT, but I don't use those much except occasionally median mode when shooting moving water.

I haven't used the motion mask as of yet - but that's an excellent suggestion. For my particular use / interest - I would be shooting well after sundown (actually after astro twilight), when its dark dark, hoping that the PS shots would provide a bit more information in terms of pulling out some landscape details - with some reduced noise. I did some experimentation last night - I drove over to a hill side and did some shots at night (10pm), and deduced some additional information. 30 second PS shots compared to a 2 minute regular exposure (within 10 seconds of each other), the PS improvements were very small. I'm thinking of putting a post together about that perhaps later this afternoon or evening. I've done a good bit of looking and there is essentially nothing out on the web on this particular use.

I've found that pixel-shifting has improved my ability to calibrate lenses, too.

That's a good piece of information!

[jbinpg]

It must be noted that unless you are in a thermally stable environment, there will always be atmospheric turbulence to deal with when using pixel shift. Especially in a hot rocky desert environment. Rawtherapees motion mask will show you just how bad this can be, even well after sundown.

[interested_observer]

The comments and information provided is excellent.

@heckflosse may be able to give some input into the situation.

I will say that even with changing light conditions I don't see the artifacts you got. Typically in Raw Therapee I choose pixel shift and then take a look at the motion mask. If it looks like movement is more than about 50 percent of the image, I don't bother with it. I then choose whichever base image is the sharpest and in your situation I would click the "equalize brightness of frames" button. At that point I save the image as a TIFF file and go from there. There are some other things you can do with pixel shift out of auto mode in RT, but I don't use those much except occasionally median mode when shooting moving water.

I'm including a motion mask down below, as it's very interesting. Yes, the lighting conditions were changing very rapidly and I do believe that is the source of the "motion" artifacts.

It must be noted that unless you are in a thermally stable environment, there will always be atmospheric turbulence to deal with when using pixel shift. Especially in a hot rocky desert environment. Rawtherapees motion mask will show you just how bad this can be, even well after sundown.

The temperature for July was really mild during the day - mid to high 90's, and in the evening was one of the nicest evening you could ever expect in the desert.

I went back and checked the times. The image was shot at 20.14, which was just after the end of civil twilight. Basically we were in "blue hour".

Here is the motion mask for the first image of the pano - below. It's interesting to note that the last 3 images of the pano have motion masks that are totally green (full frame). Thinking this through, the sun setting with respect to the first image the angle was some what acute (everything lite from the side), while the last 3 images were pretty much 180 opposite of where the sun set (or essentially front lite), so I'm thinking that the lighting change was much more pronounced.

[BigMackCam]

Here is the motion mask for the first image of the pano - below. It's interesting to note that the last 3 images of the pano have motion masks that are totally green (full frame). Thinking this through, the sun setting with respect to the first image the angle was some what acute (everything lite from the side), while the last 3 images were pretty much 180 opposite of where the sun set (or essentially front lite), so I'm thinking that the lighting change was much more pronounced.

Given the very short time between each exposure for a pixel shift image, and the presumably tiny change in lighting involved, I'm leaning towards thinking that it's atmospheric distortion at play here...

[interested_observer]

Given the very short time between each exposure for a pixel shift image, and the presumably tiny change in lighting involved, I'm leaning towards thinking that it's atmospheric distortion at play here...

That's a real possibility. I didn't bring a full weather station with me, just some Kentucky windage.