[normhead]

Since part of the thing with sensor shift is better noise performance, I'm not sure 100 ISO is the way to test it. Put it up where noise would be a problem say at 1600 ISO, and see if it cleans up the image. That's my thought anyway.

If it's a DNG a 28 MP file for the dimensions of the K-3ii, it's not going to change because you combine more images. It's probably averaging the values, not adding. The Pentax version of shifting doesn't involve an increase in detail, but more accurate colour values, at least that was my take. The pentax version is a full pixel shift, for more detail you'd have to do a half pixel shift.

[mee]

Since part of the thing with sensor shift is better noise performance, I'm not sure 100 ISO is the way to test it. Put it up where noise would be a problem say at 1600 ISO, and see if it cleans up the image. That's my thought anyway.

If it's a DNG a 28 MP file for the dimensions of the K-3ii, it's not going to change because you combine more images. It's probably averaging the values, not adding. The Pentax version of shifting doesn't involve an increase in detail, but more accurate colour values, at least that was my take. The pentax version is a full pixel shift, for more detail you'd have to do a half pixel shift.

Then why are they calling it high definition mode in the manual?

[normhead]

Then why are they calling it high definition mode in the manual?

I have no idea.....

From Imaging Resource...

Much like Olympus, Ricoh is using the image stabilization system to precisely shift the image sensor and capture multiple shots which are then merged into a single exposure. Where Olympus is performing sub-pixel shifts and using eight images to create a much higher-resolution output image than the individual source images provided by the sensor, though, Pentax is shifting by an entire pixel, producing just four exposures and outputting the merged image at the same resolution as is provided for a single exposure.

The result is an image which has full color information at every pixel location -- and thus improved resolution and a greater resistance to false color artifacts -- but only a relatively modest increase in file size. To get an idea of what full-color imaging can do compared to a Bayer-filtered camera of equal pixel resolution, we'd suggest taking a look at Sigma Foveon X3 sensor-based cameras, which have garnered quite a following for their extremely sharp, true-to-life imagery at base sensitivity.

As an added bonus, images shot in the Pixel Shift Resolution mode should also have a cleaner, tighter noise pattern. The reason for this is twofold. First, since multiple exposures are involved, noise can be averaged out across those exposures. Secondly, in a Bayer-filtered sensor, two out of three colors at each pixel location must be interpolated (read: guessed) from the values of surrounding pixels. When that happens, noise from adjacent pixels is likewise spread across their neighbors, resulting in a less film-like and blotchier, more objectionable noise pattern. With full color information at each pixel, a Pixel Shift Resolution shot's high ISO grain pattern is finer, and we're guessing easier to clean up post-capture, too.

Read all the gory details here....
Pentax K-3 II Review: Preview

Clearly the Foveon thing isn't happening, although my Foveon file so have so much sharpening applied it's hard to tell sometimes, I have some landscapes where the detail is way beyond what you can do with a bayer sensor. I don't see anything in these images that approaches that. SO that leaves noise reduction in High ISO images. For that we need high ISO image comparisons. It's somewhat disappointing for us landscape guys who always try and shoot at 100 ISO, it doesn't look like it's going to offer us much.

The only time I use high ISO is for wildlife, oops, sensor shift doesn't work on moving subjects. It would seem Pentax has laid an egg, unless I'm missing something.

[mee]

I have no idea.....

From Imaging Resource...



Read all the gory details here....
Pentax K-3 II Review: Preview

Clearly the Foveon thing isn't happening, although my Foveon file so have so much sharpening applied it's hard to tell sometimes, I have some landscapes where the detail is way beyond what you can do with a bayer sensor. I don't see anything in these images that approaches that. SO that leaves noise reduction in High ISO images. For that we need high ISO image comparisons. It's somewhat disappointing for us landscape guys who always try and shoot at 100 ISO, it doesn't look like it's going to offer us much.

Then the manual has the feature mis-labeled. Surely not.

[One3rdEV]

I may be overly optimistic, but I truly believe that for those cases where 'sufficient' lack of movement works- 'pixel shift' or 'real resolution' may be one of the most significant technologies to hit the digital still market in recent years… likely even greater in magnitude to the removal of the bayer sensor AA filter.

As I understand, by capturing full color information at each pixel, the pixel level definition increases substantially.

See the link below. The text translation does not make much sense- but pay close attention to the subtitles to each individual image and note where the enlarged examples fall within the overall scene. Note the increase in micro-detail in the enlarged magnification areas.

https://translate.google.com/translate?hl=de&sl=ja&tl=en&u=http%3A%2F%2Fwww....html&sandbox=1

---------- Post added 05-26-15 at 12:05 AM ----------

LOL- definitely went a little overboard with my first paragraph in the above post.

But seriously, do check out the examples of the above link. Shows both some impressive increase in fine details at magnification and also shows some examples of the errors that 'can' occur with subject movement (see the enlargement of the ferris wheel). Interestingly, subject movement does not always result in artifacts (as demonstrated by the motorcyclist shown in one of the magnifications). Also note the response in the trees (can observe a large area with no adverse effects, and an area with movement artifacts).

[Class A]

Thanks for posting the link as I was about to obtain it to post it here.

Pixel-shift should be called "sensel-shift" because only R, or G, or B sensels are shifted, not RGB pixels. For the same reason one should talk of Megasensels (as opposed to Megapixels) when talking about the resolution of Bayer-Array sensors. Only Foveon sensors capture RGB pixels straight away, images from Bayer-array sensors have to be demosaiced first.

That's why "Pixel-shift" goes hand in hand with omitting Bayer-AA-filters. The colour artefacts that would be introduced by not shifting the sensor as per Pentax's AA-filter simulator are eliminated in a precise manner by shifting sensels as if each sensel location contained a full RGB sensing pixel.

Hence, the "Foveon thing" is happening with Pixel-shift and noise reduction is just a welcome side effect of exactly aiming at the Foveon effect.

[RobA_Oz]

Imaging Resource hasn't tested the pixel shift technology, yet. They acknowledge that in the comments. The examples given in the linked Japanese test are probably good enough to justify the use of the "High Definition" term. It'll be interesting to see if the capture time is sufficiently short to make it useful in portraiture, or other situations where the subject isn't completely static, but where movements are slow and limited.

[Gimbal]

It’s all about details at pixel level. So how anyone can expect to see any difference while comparing two seriously downsized pictures is beyond me.

[Solaire]

Since part of the thing with sensor shift is better noise performance, I'm not sure 100 ISO is the way to test it. Put it up where noise would be a problem say at 1600 ISO, and see if it cleans up the image. That's my thought anyway.

If it's a DNG a 28 MP file for the dimensions of the K-3ii, it's not going to change because you combine more images. It's probably averaging the values, not adding. The Pentax version of shifting doesn't involve an increase in detail, but more accurate colour values, at least that was my take. The pentax version is a full pixel shift, for more detail you'd have to do a half pixel shift.

Tripod + no movement is required for pixel shift, so why would you shoot at anything other than ISO 100? Look at pixel shift as an ISO 25 option. Slower shutter, better resolution, less noise, better colors (but same DR). There are many Sony a6000 vs A7 (or even A7R) comparisons out in the wild, try looking at them at such a low resolution, I bet you won't notice the difference.

Anyway, here's the same scene wit Pixel Shift on and off. Even at fullscreen (1080p or higher depending on monitor) it's hard to see any difference. Zoom in to 100% and it's day and night. Looking at 25% or smaller, the image is already oversampled to RGB / pixel, so you won't see the ugliness of the Bayer sensor. The only difference is that now you can look at the image at 100% or print crazy big and still get pixel (or should I say dot ) perfect images.

That looks really nice. Some weird patterns on a few places but overall nice result.

[batmobile]

The difference shown above by Solaire is really quite significant and would clearly show on very large prints. One remains tight and defined, whereas the other is falling apart at 100%. Make a 40" print and nobody would be guessing which is which.... Just look at the tree leaves, signposts lettering and other fine detail.

[Rondec]

I have no idea.....

From Imaging Resource...



Read all the gory details here....
Pentax K-3 II Review: Preview

Clearly the Foveon thing isn't happening, although my Foveon file so have so much sharpening applied it's hard to tell sometimes, I have some landscapes where the detail is way beyond what you can do with a bayer sensor. I don't see anything in these images that approaches that. SO that leaves noise reduction in High ISO images. For that we need high ISO image comparisons. It's somewhat disappointing for us landscape guys who always try and shoot at 100 ISO, it doesn't look like it's going to offer us much.

The only time I use high ISO is for wildlife, oops, sensor shift doesn't work on moving subjects. It would seem Pentax has laid an egg, unless I'm missing something.

I think the question is whether you could take such a shifted image and sharpen it up more and get more dynamic range/color detail from it than from a standard Bayer sensor image. By default Foveon images are turned up to 11 and it really isn't my taste, but if there is more color detail, then you should be able to do more with the image than without the sensor shift.

[DeadJohn]

I don't have much to add. Still waiting to see if this feature can be used with Lightroom, or if we are forced to use Ricoh's software.

[MJSfoto1956]

one final test: what would the comparison be of taking four independent images (especially if handheld so that the alignment of demoisaiced pixels was random) then blending them in Photoshop? Would the IQ result be "halfway between" these two? (that is my guess)

Michael

---------- Post added 05-26-15 at 05:15 PM ----------

BTW, I feel the image posted by Solaire clearly demonstrates that this technology can be an extraordinary tool for studio and architectural work.

Michael

[One3rdEV]

Architecture, studio and ...some of those nature still photos I used to love taking with my medium format film camera mounted on a tripod (e.g. fallen leafs on an old stump saturated in fall colors, fine details and plenty of character ).