[Gimbal]

If you need better image quality for the use it could be for then you could just buy a D810 and be done with one click.

Right, one can always buy more stuff, that is a good advice.

But I was interested in the pixel shift technology as I (probably) would get it for "free" in the new K-1. That is, if I buy one. And it is not so much a question of need, but it would be fun to have.

[Cynog Ap Brychan]

It's not strange. For several reasons. It was a gimmick to start with. There was no user base request for it. It has limited use in practice. If you need better image quality for the use it could be for then you could just buy a D810 and be done with one click.

Yes, you can simply spend £2,300 for a D810... oh, and then there's a top-notch lens to go with it, say another £800-£1,000. I like taking photographs of my garden flowers in a controlled environment (my little studio) and I am amazed by the resolution and colour quality produced by pixel-shifted images. In fact, I think they are superior to shots taken with my D810 in terms of colour rendition. I don't think of pixel shift as a gimmick, by any means, just another feature to use when circumstances allow. On the other hand, I think video in a DSLR is a gimmick, so each to his/her own.

[biz-engineer]

Yes, you can simply spend £2,300 for a D810... oh, and then there's a top-notch lens to go with it, say another £800-£1,000.

Or you can do a Pentax FF upgrade (body + lenses) and spend the same amount of money. Sorry.

[RonHendriks1966]

Right, one can always buy more stuff, that is a good advice.

But I was interested in the pixel shift technology as I (probably) would get it for "free" in the new K-1. That is, if I buy one. And it is not so much a question of need, but it would be fun to have.

With the K-1 it is different again, because then you can jump over the image quality of D810 using pixel shift. I don't expect it to become popular.

[Cynog Ap Brychan]

Or you can do a Pentax FF upgrade (body + lenses) and spend the same amount of money. Sorry.

I think you've missed the point here, which was about pixel shift on the K-3 II. Its implementation on the K-1 should result in much better files than the D810 is capable of under the right circumstances.

[biz-engineer]

I think you've missed the point here, which was about pixel shift on the K-3 II. Its implementation on the K-1 should result in much better files than the D810 is capable of under the right circumstances.

Yes, I got your point. I want to cut all Pentaxforum discussions about this pixel shift, short: JPEG is 8 bits deep, so, beyond 48dB of SFDR, all information goes down the drain once you compress you raw image to JPEG for storage or printing. At base ISO, you already got about 42 dB of SFDR with your K-3, when you go FF, you get 6 more dB (8 bits effective), and then you see no difference at all, except if you increase the ISO, BUT, pixel shift = tripod = 100 ISO. In other words, nobody need pixel shift on a full frame camera, except for marketing reasons.

Best regards.

[DavidOliver]

[...] super resolution can be achieved by manually combining standard files taken with just about any camera.

By shifting the camera one pixel between every shot?

I think Adam may have been referring to hand-shifted images composited (and aligned) with Photoshop or similar, as per this article:

When I first tried this technique, I tried using a tripod, taking a single image, and then tapping the tripod just the slightest bit to “shift” the “sensor” (i.e. the entire camera) before taking another image. Rinse and repeat. The problem with this technique is that, while it works, it takes a long time to take a lot of images and it’s difficult to keep the movements small enough.


I found it’s actually much easier and more practical to just set your camera to continuous burst mode and rattle off a bunch of handheld images. We don’t even need to try to intentionally move the camera to emulate the sensor shift as the natural instability of our hands is enough to make make the small shifts needed for superresolution compositing.

It's pretty cool that this can be done without a tripod. I don't know if the results are as good as pixel shifting.

[Adam]

I think Adam may have been referring to hand-shifted images composited (and aligned) with Photoshop or similar, as per this article:

Exactly

[Gimbal]

Yes, I got your point. I want to cut all Pentaxforum discussions about this pixel shift, short: JPEG is 8 bits deep, so, beyond 48dB of SFDR, all information goes down the drain once you compress you raw image to JPEG for storage or printing. At base ISO, you already got about 42 dB of SFDR with your K-3, when you go FF, you get 6 more dB (8 bits effective), and then you see no difference at all, except if you increase the ISO, BUT, pixel shift = tripod = 100 ISO. In other words, nobody need pixel shift on a full frame camera, except for marketing reasons.

Best regards.

That is a weird statement.
I will not comment about the ”need” as that is up to the user, but the pixel shift will of course improve the picture in the same way as it does on the K3 at 100% crop. There is nothing magic going on at 100% crop on a full frame that doesn’t happen on APS-c. At pixel level the sensors are the same.

[Cynog Ap Brychan]

I will not comment about the ”need” as that is up to the user, but the pixel shift will of course improve the picture in the same way as it does on the K3 at 100% crop. There is nothing magic going on at 100% crop on a full frame that doesn’t happen on APS-c. At pixel level the sensors are the same.

Thank you Gimbal for confirming what I was thinking. Despite biz-engineer's explanation of the underlying physics (of which I know little) I couldn't fathom how pixel shift would lead to a better image on an APS-C sensor and not on a full-frame one. After all, the camera is taking a RBG reading at each pixel site location rather than interpolating the colours from surrounding pixels, somewhat in the manner of a Foveon sensor, though of course that does it in one take. I think biz is talking about noise levels, though I'm not sure about this either. By making four exposures, the camera is reading information from four times more light, although some of that information is probably discarded when amalgamating the readings to make the final RAW file. Would that not result in lower noise levels? Perhaps someone more knowledgeable than I would care to elaborate on this. I only know that when I used pixel shift for the first time on the K-3 IIs, I was blown away by the clarity and colour rendition which, to me, looked better than my D810.

[Rondec]

Yes, I got your point. I want to cut all Pentaxforum discussions about this pixel shift, short: JPEG is 8 bits deep, so, beyond 48dB of SFDR, all information goes down the drain once you compress you raw image to JPEG for storage or printing. At base ISO, you already got about 42 dB of SFDR with your K-3, when you go FF, you get 6 more dB (8 bits effective), and then you see no difference at all, except if you increase the ISO, BUT, pixel shift = tripod = 100 ISO. In other words, nobody need pixel shift on a full frame camera, except for marketing reasons.

Best regards.

I guess I would say a couple of things. First of all, it is hard to say what anyone "needs." Most people don't "need" full frame and who really "needs" medium format cameras. And yet they buy them. Whether it is a need or not, people do pursue better image quality and I think it is safe to say that this is one way to improve image quality -- better color depth and resolution. If Pentax came out with a Foveon sensor full frame camera that offered this sort of thing in a single image, I don't think people would cast aspersions on it in this way. The second thing, is that the K3 II version is an initial implementation. I wouldn't foresee a version that ever worked without a tripod, but certainly future iterations of this feature will have less artifacts from subject movement.

In the end, pixel shift is a tool. It will be useful to some folks and not to others, but it certainly doesn't hurt that Pentax includes it on their cameras and I, for one, am glad that the K1 will have it so I can experiment with it in the future.

[rawr]

PS captures the actual R, G and B for each pixel and no interpolation.

Which, personally, I think adds as much to the value of pixel-shift as the resolution boost, which gets most of the attention.

So pixel shift delivers three things:

- greater resolution;
- improved colour reproduction;
- noise reduction.

I certainly look forward to seeing it at work in the K-1. What's not to like?

I wonder if the K-1 pixel-shift will be exactly the same pixel-shift as seen in the K-3II, or some manner of pixel-shift v2.0 ?

[RonHendriks1966]

That is a weird statement.
I will not comment about the ”need” as that is up to the user, but the pixel shift will of course improve the picture in the same way as it does on the K3 at 100% crop. There is nothing magic going on at 100% crop on a full frame that doesn’t happen on APS-c. At pixel level the sensors are the same.

This is not true for all sensors and pixels. Only if the full frame sensor would be cut out of the same wafer as the K-3ii is. Other wise sensors and pixels can be very different in quality.

[Gimbal]

This is not true for all sensors and pixels. Only if the full frame sensor would be cut out of the same wafer as the K-3ii is. Other wise sensors and pixels can be very different in quality.

Right, if anyone thought that I meant that all sensors ever made are equal I apologies.

[pinholecam]

I get the impression that lots of shooters here on PF are very 'traditional' (maybe older folkds too).
So the style of shooting is still old style 'straight shots'.
Less digital blending and photoshop "trickery".
Don't get me wrong, I certainly believe there are dozen ways to skin the cat, just citing that the preference of approach does not naturally lend itself to exploring the capabilities of pixel shift.

So in order to get a tech savvy shooter who actually buys a K3II and bothers to learn/explore the possibilities of pixel shift, furthermore, to actually start talking/posting about it.
All under the shadow of a FF coming (and spending on it over the K3II)
I think that chance is lower.

Thank you Gimbal for confirming what I was thinking. Despite biz-engineer's explanation of the underlying physics (of which I know little) I couldn't fathom how pixel shift would lead to a better image on an APS-C sensor and not on a full-frame one. After all, the camera is taking a RBG reading at each pixel site location rather than interpolating the colours from surrounding pixels, somewhat in the manner of a Foveon sensor, though of course that does it in one take. I think biz is talking about noise levels, though I'm not sure about this either. By making four exposures, the camera is reading information from four times more light, although some of that information is probably discarded when amalgamating the readings to make the final RAW file. Would that not result in lower noise levels? Perhaps someone more knowledgeable than I would care to elaborate on this. I only know that when I used pixel shift for the first time on the K-3 IIs, I was blown away by the clarity and colour rendition which, to me, looked better than my D810.

I think you can summarize it as :
"I have the cameras and there is a difference! "