[Chris Hopton]

I have read many reviews which insists that pixel-shift mode can be used only when the camera is secured to a sturdy tripod. This makes perfect sense to me since I cannot believe that any human being can hold a camera so still as to match a pixel-width in movement as four shots are taken and stitched together. Nevertheless, initially more through accident than design, I have discovered I can take crystal clear street photos in pixel shift mode, handholding the K3II camera, and with shutter speeds at a modest 1/60th of a second and ISO at 100. Yesterday I took 25 handheld high res shots at Orford in Suffolk and every one of them was sharp and detailed although one (of the cemetery) did show on the JPEG image some downward motion blur (which was not at all apparent in the DNG composite shot). I've now emailed RICOH to ask them how this is possible. Have they undersold the K3II's full potential? The weather yesterday at Orford was also quite blustery which should have compounded the problem of getting any usable shots.

I have failed to find any support on the web for handholding the camera for pixel-shift: everybody rules it out as an impossibility if you want to avoid "zipper artefacts". I wouldn't believe it either if I hadn't seen firsthand the results in DXO Labs of the DNGs produced by my K3II with the 31mm limited edition lens.

I'd attach samples if the file sizes were not so restricted.

Has anybody else had success with handholding the camera whilst using pixel-shift mode?

[Rondec]

I have read many reviews which insists that pixel-shift mode can be used only when the camera is secured to a sturdy tripod. This makes perfect sense to me since I cannot believe that any human being can hold a camera so still as to match a pixel-width in movement as four shots are taken and stitched together. Nevertheless, initially more through accident than design, I have discovered I can take crystal clear street photos in pixel shift mode, handholding the K3II camera, and with shutter speeds at a modest 1/60th of a second and ISO at 100. Yesterday I took 25 handheld high res shots at Orford in Suffolk and every one of them was sharp and detailed although one (of the cemetery) did show on the JPEG image some downward motion blur (which was not at all apparent in the DNG composite shot). I've now emailed RICOH to ask them how this is possible. Have they undersold the K3II's full potential? The weather yesterday at Orford was also quite blustery which should have compounded the problem of getting any usable shots.

I have failed to find any support on the web for handholding the camera for pixel-shift: everybody rules it out as an impossibility if you want to avoid "zipper artefacts". I wouldn't believe it either if I hadn't seen firsthand the results in DXO Labs of the DNGs produced by my K3II with the 31mm limited edition lens.

I'd attach samples if the file sizes were not so restricted.

Has anybody else had success with handholding the camera whilst using pixel-shift mode?

I would be interested for you to compare them in Raw Therapee -- that is to look at an individual frame of the four captured with a pixel shift burst followed by looking at the combination of the four with motion correction and then finally to look at the motion mask for the combined image.

I think there is a strong placebo effect with pixel shift images. We think we should see a difference and so we do, but the reality is that, even with the scene is still and the camera is on a tripod, there is little difference between a single shot capture and the pixel shift image. You have to push shadows and do a lot of sharpening in order to really perceive a difference.

I don't own a K-3 II, but my experience with the K-1, K-1 II and K-3 III used with Raw Therapee is that pixel shift isn't very effective (if at all) hand held.

[Bob 256]

Technically, you can't achieve the pixel shift effect handheld unless you're a stone statue. It's virtually impossible to match each and every pixel to its counterpart if there's the least bit of motion between the pixel shift exposures. Many scenes contain moving objects (water, birds, etc.) and pixel shift cheats on those to prevent artifacts from occurring, but if the whole frame shifts, all bets are off.

[microlight]

I tried to do this when I got my K-3II in 2016 but after a few attempts came to the conclusion that the shots were really no better than standard shots taken alongside for reference - and the files were four times the size! 😉

[OrchidJulie]

Hi, Chris...welcome to PentaxForums!

I am moving this thread to the K3 section, as that seems a better home for it. You'll get the attention of more K3 users there.

Moderator

[mlag]

in my latest attempt the wind on tree leaves and flowing water ruined my landscape pictures ...
yes , it was handheld and some parts were ok...but too many parts not, especially were one could suspect natural movements

[rpjallan]

If you are using DXO software you are probably only seeing one of the images because, as far as I am aware, none of their software supports Pentax pixel shift images.

[tduell]

If you are using DXO software you are probably only seeing one of the images

This too had crossed my mind.
You could get dcrawps and using your PS raw image as input, dcrawps should tell you whether it is a PS image and provide a processed (combined) output, complete with a display of the artefacts resulting from movement, if asked to do so.
Invoking dcrawps without arguments will display all the options.
A very useful tool if playing around with pixelshift.

Cheers,
Terry

[Chris Hopton]

I am using DXO Photolab 5.8.0. The DNG files I am working on vary between 100Mb and 130Mb: so, these are all vastly greater than the normal DNG files produced by a single shot. I've attached the metadata from one of the shots so you can see the difference. I posted only because I am fully aware that I should not be getting the crystal-clear images that I am. Zooming in to the images immediately shows the detail that's missing from the ordinary shot. Since the camera composits the four images within the camera body and produces one large DNG I didn't anticipate that the DXO software would have an issue with it.

[Chris Hopton]

Hi Bob, That was the very point I was making at the start of my post. I want to uncover why, given the impossibility of holding a camera pixel-still, I'm getting large DNG files (on average 120Mb in size) which show far greater detail in depth of field than the standard shots (which are usually 29Mb in size). I posted only because I was seeing the totally unexpected. Hence the reason too why I emailed Ricoh to see if they could explain.

[Chris Hopton]

Hi Rondec,

For there to be a placebo effect, I'd have to be expecting to see a vast difference. I really didn't. I took the photos in pixel-shift mode, handheld, fully expecting to see blurry images with massive and obvious movement. I thought I might get some indicators of how I could improve my posture for better handholding shots. When I ended up with really sharp images, with great colour and great depth-of-field (but huge file sizes) I thought I'd contact RICOH and post my findings. The images look very different from anything I have formerly shot on a K3. I have shown the photos to others who think the images look more like something shot on full-frame or Medium format. I started my post by highlighting my expectation that handholding pixel-shift mode would simply not work; it's only because I saw what I was NOT expecting that I posted. I'm not familiar with RAWtherapee but am not sure it would help me since the SD card shows only the single huge DNG from the four images composited together in the camera.

[pschlute]

This is interesting. Perhaps you could post a link to the PS raw file itself via dropbox or the like.

[syyrmb]

I believe it's just that more sharpening was applied to the "handheld pixel-shift" JPEGs you're seeing. This behavior makes sense since in normal circumstances, pixel shift allows for more aggressive sharpening due to lessen overall noise level in the image. And if you disable motion correction you will see that there will be many artifacts in the image (not so much in the lower frequency areas like blue sky, white wall, bokeh, etc. though), this can be demonstrated more prominently if you check the "show motion correction mask" checkbox in Rawtherapee. You'll see that most (if not the entire image) is covered by green mask, meaning most of the image is moving and not "pixel shifted". I've done this experiment myself and this is what I found (and I didn't even mention the jello effect of the electronic shutter which could cause pixel value change when there is movement). Also one of the developer of Rawtherapee concluded that any movement doesn't cooperate well at all with pixel shift and demonstrated his point with an image that has one pixel of movement in the 4 images' sequence and that exact pixel shows movement artifact. All in all, pixel shift is not gonna work without tripod like the traditional wisdom suggests, if you think it works, it usually because of some software post-processing (sharpening) or the software not able to correctly decode pixel shift image thus only shows the first image in the 4 pixels shift sequence, which makes no difference to a non-pixel-shifted image (only softwares that supports pixel shift that I know are Adobe Lightroom (doesn't support motion correction though), Silkypix (support motion correction so it's better than LR, I guess), Pentax DCU (also support MC and the MC algorithm is different from Silkypix for some reason, MC works better than Silkypix in my very limited test, yes, test, not tests), Rawtherapee (many advanced MC options, though Auto MC works quite well out of the box) and DCRaw which is a command line tool).

[tduell]

Since the camera composits the four images within the camera body and produces one large DNG I didn't anticipate that the DXO software would have an issue with it.

It's been a while since I've used pixelshift, so I may be mistaken, but I think that a pixelshift raw file (DNG or PEF) produced by the camera is simply a container with the four individual images, that then need to be post processed to combine those four images into the single image result.
If your post processing software doesn't have the tools to do that it will probably only work with one of the four images.
Hopefully someone will correct me if I've misremembered on this.

Cheers,
Terry

[Rondec]

Hi Rondec,

For there to be a placebo effect, I'd have to be expecting to see a vast difference. I really didn't. I took the photos in pixel-shift mode, handheld, fully expecting to see blurry images with massive and obvious movement. I thought I might get some indicators of how I could improve my posture for better handholding shots. When I ended up with really sharp images, with great colour and great depth-of-field (but huge file sizes) I thought I'd contact RICOH and post my findings. The images look very different from anything I have formerly shot on a K3. I have shown the photos to others who think the images look more like something shot on full-frame or Medium format. I started my post by highlighting my expectation that handholding pixel-shift mode would simply not work; it's only because I saw what I was NOT expecting that I posted. I'm not familiar with RAWtherapee but am not sure it would help me since the SD card shows only the single huge DNG from the four images composited together in the camera.

I wasn't really speaking of you in particular with the placebo effect.

I am not familiar with DXO Software, but I use Lightroom 5.6. Now, Lightroom sort of can process pixel shift images from the K-1. It does so without motion correction and so there are artifacts all over the place. On the other hand, with K-1 II images, it simply takes the first raw image of the four image stack and processes that. You see no artifacts, but that is simply because it isn't really combining the images. That is what many software packages do with pixel shift images and I think that is what DXO software does.

Raw Therapee is free and it has the best motion correction software -- most flexible and able to choose which is the base image out of the four. That's why I suggested trying it to see how that looks at it.