[Sperdynamite]

So far there were two approaches to Pixel Shift from my understanding. One was the Pentax mode which we call know and love. It essentially uses pixel shift to get around the need for bayer demosaicing. You get the full possible image quality of the sensor. The other was the Olympus super res mode, which increased the pixel count. Given that 36mp is still a lot, I prefer the Pentax method.

I'm not sure if Pentax users noticed but two new Pixel Shift capable cameras have been released. The Sony A7R IV, and the Panasonic S1R. These models combine the Pentax and Olympus method to get around bayer demosiacing and increase resolution. I believe the Sony makes 8 captures, and promises 240mp.

Now I don't get too excited about MP increases for practical photography. However I primarily use pixel shift for scanning film, a task to which it is perfectly suited. When you go to formats like 1:1 you pay a significant resolution price with single capture scans. You get around this by taking multiple scans of the film in sections and stitching them together. 6 captures of a 6x6 negative yields close to 4000ppi. The super resolution mode that the Sony and Panasonic may reduce any need for stitching, which is time consuming and can be challenging at times.

I guess I am just hoping that Pentax can upgrade the Pixel Shift mode in a firmware update to offer more captures in pixel shift mode to increase resolution.

[biz-engineer]

You get around this by taking multiple scans of the film in sections and stitching them together. 6 captures of a 6x6 negative yields close to 4000ppi.

With pixel shift super resolution, hopefully you can find a lens that resolve 4000ppi over a 24x36mm sensor. 4000ppi, that's 787 lp/mm, sharpest photographic lens resolve about 70 lp/mm in the center only.

[Trickortreat]

The whole combination will still resolve a bit more than than without SR, regardless of lens used. So it would be beneficial and future proof… besides if theres a 24MP FF down the road it might do new lenses justice.

[Fogel70]

With pixel shift super resolution, hopefully you can find a lens that resolve 4000ppi over a 24x36mm sensor. 4000ppi, that's 787 lp/mm, sharpest photographic lens resolve about 70 lp/mm in the center only.

Why would you need that? Each captured image is still standard 36MP.
The increased resolution happen without any optics involved.

[biz-engineer]

The increased resolution happen without any optics involved.

The OP isn't asking for 4 x pixel shift, Pentax already features that. He's asking for super resolution with 16 x frames or more like Olympus, which is limited by the optics.

[Fogel70]

The OP isn't asking for 4 x pixel shift, Pentax already features that. He's asking for super resolution with 16 x frames or more like Olympus, which is limited by the optics.

Why is it that more demanding when each of the 16 frames are the same old 36 MP captures with standard resolution?
You get the increased resolution when you add all them together in software.

[beholder3]

Why is it that more demanding when each of the 16 frames are the same old 36 MP captures with standard resolution?
You get the increased resolution when you add all them together in software.

Because you only get additional information from a higher pixelcount by either adding more "area" (that is the typical stichted panorama, which this is not) or you get more information from he same area (and this inevitably means more information for a tiny area must come through the lens).

Pentax pixelshift raised the effective resolition from 36 Mpx Bayer to the equivalent of about 50 MPx Bayer. And yes, by Pixelpeeping you can see a difference if you search for it. But even this is beyond what the vast majority of people viewing any image will ever acknowledge (see all the threads on the topic).
Raising it even more means that - even if you use a stopped down lens that is capable of resolving enough - even fewer people will ever see a difference.

[biz-engineer]

Lenses optical resolving power, for a good lens, is usually around 50 line pairs per mm (MTF charts test pattern using 30 lp/mm and 10lp/mm give evolution of the contrast value of line pair from the center of the lens to the edges), to sample 50 lp/mm at double nyquist, is 200 pixels per mm on the sensor (alternated 100 pixels red, 100 pixels green, 100 pixels blue), that's 5um pixel pitch. The K1 sensor has 204 pixels / mm, that just about right even for a sharp prime lens. Higher pixel density grabs the lower that 50% contrasts detail in the image projected on the sensor. When the pixel density increases beyond 200 pixels/mm, the resolution of the digital image doesn't increase as fast as pixel density, at some point more pixels don't bring anything except the making of bigger files. Decreasing pixel pitch below 5um follow bring diminishing improvement in resolution but the size of uncompressed raw files increases proportionally to the total pixel count. A center crop of the image benefits more from higher pixels density because lenses optical resolution is better in the center, but no advantage in bumping up pixel density at the edges of the image frame.

[Fogel70]

Because you only get additional information from a higher pixelcount by either adding more "area" (that is the typical stichted panorama, which this is not) or you get more information from he same area (and this inevitably means more information for a tiny area must come through the lens).

Are you saying that if you capture an image and shift the sensor half a pixel and capture a second image, the second image you capture will have less resolution than the first?

Pentax pixelshift raised the effective resolition from 36 Mpx Bayer to the equivalent of about 50 MPx Bayer. And yes, by Pixelpeeping you can see a difference if you search for it. But even this is beyond what the vast majority of people viewing any image will ever acknowledge (see all the threads on the topic).

Raising it even more means that - even if you use a stopped down lens that is capable of resolving enough - even fewer people will ever see a difference.

Pentax pixel shift mainly increase color accuracy, which also improve resolution some. But it is still 36MP data.

[biz-engineer]

Pentax pixel shift mainly increase color accuracy, which also improve resolution some. But it is still 36MP data.

I think what beholder meant is: if you take a 50Mpixel sensor and do bayer interpolation, you get similar image details as if you take a 36Mpixels sensor with pixel shift instead of a bayer interpolation.
Since the light collected in pixel shift mode is 4 x the light collected by a single frame, you also get half the noise with PS. So pixel shift 36Mp at ISO100 is a bit like shooting a single 50Mp bayer interpolated frame at ISO 25.

[Trickortreat]

The OP isn't asking for 4 x pixel shift, Pentax already features that. He's asking for super resolution with 16 x frames or more like Olympus, which is limited by the optics.

Yes, I know, but even if the lens cant handle the full super res resolution there would be an increase in final image resolution nonetheless...

[Rondec]

Are you saying that if you capture an image and shift the sensor half a pixel and capture a second image, the second image you capture will have less resolution than the first?


Pentax pixel shift mainly increase color accuracy, which also improve resolution some. But it is still 36MP data.

I guess the question is what you do when you are capturing more color detail. With pixel shift you should get something more like a Foveon level of color information. Sigma says that the SD 14 sensor has 14 megapixels when the reality is that it has 4.7 megapixels for each color. The end result is something better than 5 megapixels, but not as good as 14 megapixels either.

With pixel shift, you definitely get extra resolution and comparing images from the Canon 5Ds to a K-1 pixel shift, you find that the pixel shift images seem to have more detail in them, even though they have fewer megapixels.

[biz-engineer]

Yes, I know, but even if the lens cant handle the full super res resolution there would be an increase in resolution nonetheless...

No sure, it depends... I've played with pixel shift of the K1 in order to find for myself when I get a benefit and when I don't get a benefit. For cases where I don't get a benefit it is more simple to take a single frame with electronic shutter.
I've tried pixel shift at various lens apertures with newspaper on the wall as a test chart, using the 100mm macro , focused with 8x mag. in LV.
At aperture f11, I get zero improvement of sharpness with PS vs a single shot ES.
At aperture f8, I get very little improvement of sharpness with PS vs single shot ES.
At aperture f5.6, I can see a bump in sharpness using PS compared to a single shot with ES.
At aperture f4, I can see super sharp pixels in the center but not at the edges when using PS , compared to ES baseline.
So , for example, I never use PS when shooting a landscape at f11 or f16 and ISO100, because.. for me the noise is already very low at ISO 100 and PS doesn't bring me more sharpness at f11 or f16 (lens diffraction).
However, for a macro flat subject, I set the crop mode on K1, get as close as possible, lens aperture at f5.6, and there I get wowed by the detail and color micro-contrast.

So given what I can see of improvements condition with a 4 x pixel shift on the K1, I don't think the 16 x half pixel shift on the Sony A7R4 will be having a benefit enough to bother storing 2Gb files and processing them.

[photoptimist]

I guess the question is what you do when you are capturing more color detail. With pixel shift you should get something more like a Foveon level of color information. Sigma says that the SD 14 sensor has 14 megapixels when the reality is that it has 4.7 megapixels for each color. The end result is something better than 5 megapixels, but not as good as 14 megapixels either.

With pixel shift, you definitely get extra resolution and comparing images from the Canon 5Ds to a K-1 pixel shift, you find that the pixel shift images seem to have more detail in them, even though they have fewer megapixels.

The "megapixel" specs for Bayer filter cameras are a bit of a cheat. The Canon 5Ds may claim 50 MPix but it is really a sensor that combines a 12.5 Mpix red-sensitive sensor array, 12.5 Mpix bluer-sensitive sensor array, and a 25 Mpix red-sensitive sensor array. Similarly, the K-1 in normal single-shot mode is only 9 Mpix in red, 9 MPix in blue, and 18 MPix in green. Demosaicing tries to fill in the missing data but it is akin to "digital zoom" touted on cheap P&S cameras.

Pixel shift moves those red, blue, and green arrays around in space to fill in all the missing data and get a true 36 MPix in red, green, and blue. As such, a K-1 PS image has almost 3X pixel count in red and blue as the 5Ds and almost 1.5X the pixel count in green.

Super-resolution of the kind offered by Olympus, Hasselblad, and Sony is a whole other level of technology. It goes beyond simple PS by taking shots with various half-pixel shifts and then mathematically analyzes the pixel-to-pixel and frame-to-frame differences. This analysis estimates or reconstructs the image at much higher sampling resolution than the native sensor's resolution. Moreover, super-resolution is not limited to the resolving limit of the lens although the reconstruction of the super-resolution image will be noisy if the lens cannot resolve details are the resolution of the super-resolution. Thus, super-resolution also takes advantage of the very low-noise, high-DR performance of modern sensors.

[beholder3]

Are you saying that if you capture an image and shift the sensor half a pixel and capture a second image, the second image you capture will have less resolution than the first?

No. Where did I write that?

Pentax pixel shift mainly increase color accuracy, which also improve resolution some. But it is still 36MP data.

As I said: the equivalent of +40% detail resolution with a single shot.

Matter of fact it turns the original 4x 9 MPx + interpolation guesswork resolution into 36 MPx resolution.