Fuji should look to the Sensor shift in the P K3-II

Well worth a look at the DPReview Test chart. What a dramatic improvement in resolution and colour purity!

The chart blows away practicaly all the 30Mp+ sensors out there, FF or not. I would say it has similar resolution as their MF camera now. What it does also show up is the poor edge perfromance of the PK lens used! Also just read the envious comments from the CanNikon users.

I wonder if the XTrans chip, on a suitible sensor shift could do something similar? Fuji I'm sure you are thinking of this for the X-PRO2

Yes I know it has to be tripod mounted to achieve this, but when you need this level of image resolution you will need a tripod whatever the camera. So I don't see this as a major issue.

Well done Pentax!

What no one seems to point out is that you have been able to accomplish essentially the same thing, a high resolution image with multiple exposures with image stitching for years and years. Once a tripod comes into play nothing is convenient. And you are not limited by the programming in the camera. You want 100 MP? Fine.

I think the multi-shot techniques based on shift-mechanisms are one of the most exciting recent developments in the camera world. With procedures that involve single-pixel shifts, one gets full color information for every pixel, like from a Foveon sensor. That would abolish requirements for demosaicing, thus removing a big computational hurdle from image processing.

Unfortunately, I also think that the latter is the reason why we probably won't see it in the Fuji cameras anytime soon. Fuji simply has too much riding on the X-Trans CFA and the associated demosaicing procedures. Throwing all this away would be a step too big for Fuji to take - at least at this point.

I have been trying to replicate high-resolution methods by shooting several, slightly shifted images. It's a pain to build a proper pipeline for assembling a final image. In addition, the need for demosaicing doesn't go away.

If Fuji won't implement shift technologies while others do, it would be a reason for me to jump ship, because my preferred type of photography would greatly benefit from increased resolution and color fidelity.

win39 said:

What no one seems to point out is that you have been able to accomplish essentially the same thing, a high resolution image with multiple exposures with image stitching for years and years. Once a tripod comes into play nothing is convenient. And you are not limited by the programming in the camera. You want 100 MP? Fine.

Click to expand...

And don't forget it's useless on everything but shooting inert objects in studio (even landscapes are affected by wind) so you are right ! Might as well use focus stacking and get insane resolution and detail
--

VictorTrasvina said:

VictorTrasvina said:

What no one seems to point out is that you have been able to accomplish essentially the same thing, a high resolution image with multiple exposures with image stitching for years and years. Once a tripod comes into play nothing is convenient. And you are not limited by the programming in the camera. You want 100 MP? Fine.

Click to expand...

And don't forget it's useless on everything but shooting inert objects in studio (even landscapes are affected by wind) so you are right ! Might as well use focus stacking and get insane resolution and detail

Click to expand...

It is by no means straightforward to create high-resolution images from multiple shots. One way to achieve that is by using a longer focal length and stitching together images. But then, DoF would potentially be a problem.

One can also shift images and assemble higher-resolution versions from those. That is very difficult, though; there are no good tools as far as I know. And connecting the tools that cover certain aspects (3D alignment and averaging of images) is not a trivial task.

besides, I would contend that taking the images required for the mentioned techniques takes more time than a quick burst of three images that are automatically shifted by 1 pixel. Even for non-stationary objects, the method can be beneficial, depending of course on the amount of movement and exposure times.

In summary, in-camera shift technologies aren't easy to beat, if they can beat at all (IMHO).

It's the ability to add RGB info to every pixel which, like Fovion, gives you both a resolution increase allied to colour purity which makes the difference. Unlike Olympus there are no half pixel shifts here

As the sensor has a different colour mask it could do a similar shift, possibly with movement in one axis rather than two

I don't know if this is process is subject to copywrite

PG Thomas said:

It's the ability to add RGB info to every pixel which, like Fovion, gives you both a resolution increase allied to colour purity which makes the difference. Unlike Olympus there are no half pixel shifts here

As the sensor has a different colour mask it could do a similar shift, possibly with movement in one axis rather than two

I don't know if this is process is subject to copywrite

Click to expand...

Getting the full color information on every pixel from three images requires a new CFA; neither the traditional Bayer, nor the X-Trans CFA could be used. It would need to be a simple pattern:

RGB
GBR
BRG

Unfortunately, that would also mean that the pattern couldn't be used for single imaging. It seems like there would either have to be a special sensor that is designed solely for multi-shot techniques (in which case, one could also simply use a color wheel on top of a monochrome sensor), or one would need to use more than three images.

PG Thomas said:

Well worth a look at the DPReview Test chart. What a dramatic improvement in resolution and colour purity!

The chart blows away practicaly all the 30Mp+ sensors out there, FF or not. I would say it has similar resolution as their MF camera now. What it does also show up is the poor edge perfromance of the PK lens used! Also just read the envious comments from the CanNikon users.

I wonder if the XTrans chip, on a suitible sensor shift could do something similar? Fuji I'm sure you are thinking of this for the X-PRO2

Yes I know it has to be tripod mounted to achieve this, but when you need this level of image resolution you will need a tripod whatever the camera. So I don't see this as a major issue.

Well done Pentax!

Click to expand...

I have had my eyes on the K3II for a while, very exiting tecnology. Unfortunately K3II users themselves are saying that in real life samples they are not seeing any difference compared to their K3.

That's why right now the K3 is selling like hot cakes

nixda said:

VictorTrasvina said:

nixda said:

What no one seems to point out is that you have been able to accomplish essentially the same thing, a high resolution image with multiple exposures with image stitching for years and years. Once a tripod comes into play nothing is convenient. And you are not limited by the programming in the camera. You want 100 MP? Fine.

Click to expand...

And don't forget it's useless on everything but shooting inert objects in studio (even landscapes are affected by wind) so you are right ! Might as well use focus stacking and get insane resolution and detail

Click to expand...

It is by no means straightforward to create high-resolution images from multiple shots. One way to achieve that is by using a longer focal length and stitching together images. But then, DoF would potentially be a problem.

One can also shift images and assemble higher-resolution versions from those. That is very difficult, though; there are no good tools as far as I know. And connecting the tools that cover certain aspects (3D alignment and averaging of images) is not a trivial task.

besides, I would contend that taking the images required for the mentioned techniques takes more time than a quick burst of three images that are automatically shifted by 1 pixel. Even for non-stationary objects, the method can be beneficial, depending of course on the amount of movement and exposure times.

In summary, in-camera shift technologies aren't easy to beat, if they can beat at all (IMHO).

Click to expand...

I think you overrate this new technology...

The main problem is that it behaves as if you had a higher resolution sensor but the resolution of the lens remains the same and this is in most cases the limiting factor. You can not make miracles except if you have a more than excellent lens !

In quality, you beat if you use the technic of stitching multiple shots with a longer focal length. The final real resolution will be much higher.

I can make a real 100mp resolution image by stitching. No problems. But this is not possible with sensor shift, the real resolution will be much lower than the number of pixels.

I don't think this is revolutionary in fact, it has no effect on the optical quality. . This is just a nice to have.

Just give me more pixels (16mp is low), it will fullfill my needs.

Christof21 said:

nixda said:

VictorTrasvina said:

Member said:

What no one seems to point out is that you have been able to accomplish essentially the same thing, a high resolution image with multiple exposures with image stitching for years and years. Once a tripod comes into play nothing is convenient. And you are not limited by the programming in the camera. You want 100 MP? Fine.

Click to expand...

And don't forget it's useless on everything but shooting inert objects in studio (even landscapes are affected by wind) so you are right ! Might as well use focus stacking and get insane resolution and detail

Click to expand...

It is by no means straightforward to create high-resolution images from multiple shots. One way to achieve that is by using a longer focal length and stitching together images. But then, DoF would potentially be a problem.

One can also shift images and assemble higher-resolution versions from those. That is very difficult, though; there are no good tools as far as I know. And connecting the tools that cover certain aspects (3D alignment and averaging of images) is not a trivial task.

besides, I would contend that taking the images required for the mentioned techniques takes more time than a quick burst of three images that are automatically shifted by 1 pixel. Even for non-stationary objects, the method can be beneficial, depending of course on the amount of movement and exposure times.

In summary, in-camera shift technologies aren't easy to beat, if they can beat at all (IMHO).

Click to expand...

I think you overrate this new technology...

The main problem is that it behaves as if you had a higher resolution sensor but the resolution of the lens remains the same and this is in most cases the limiting factor. You can not make miracles except if you have a more than excellent lens !

Click to expand...

The inherent sensor resolution is exactly the same. Imagine you'd be using a monochrome version of it or a monochrome film simulation. That version already has the full "color" resolution, and the lenses are more than capable of supplying that resolution.

Member said:

In quality, you beat if you use the technic of stitching multiple shots with a longer focal length. The final real resolution will be much higher.

I can make a real 100mp resolution image by stitching. No problems. But this is not possible with sensor shift, the real resolution will be much lower than the number of pixels.

Click to expand...

I often run into DoF issues when using longer focal lengths. Also, I do appreciate not having to get my 3D panorama setup out all the time and not having to stitch images. All that I think would be quite welcome in general.

But basically, you are arguing against any kind of higher-resolution sensor, because one can always stitch together images, even if the sensor had only 1MP.

Member said:

I don't think this is revolutionary in fact, it has no effect on the optical quality. . This is just a nice to have.

Click to expand...

I don't know what you mean by "optical quality", but the image quality would be improved tremendously. Just compare the output of a Foveon sensor to that of a traditional sensor with the same resolution. It's quite easily visible.

nixda said:

Christof21 said:

nixda said:

VictorTrasvina said:

nixda said:

What no one seems to point out is that you have been able to accomplish essentially the same thing, a high resolution image with multiple exposures with image stitching for years and years. Once a tripod comes into play nothing is convenient. And you are not limited by the programming in the camera. You want 100 MP? Fine.

Click to expand...

And don't forget it's useless on everything but shooting inert objects in studio (even landscapes are affected by wind) so you are right ! Might as well use focus stacking and get insane resolution and detail

Click to expand...

It is by no means straightforward to create high-resolution images from multiple shots. One way to achieve that is by using a longer focal length and stitching together images. But then, DoF would potentially be a problem.

One can also shift images and assemble higher-resolution versions from those. That is very difficult, though; there are no good tools as far as I know. And connecting the tools that cover certain aspects (3D alignment and averaging of images) is not a trivial task.

besides, I would contend that taking the images required for the mentioned techniques takes more time than a quick burst of three images that are automatically shifted by 1 pixel. Even for non-stationary objects, the method can be beneficial, depending of course on the amount of movement and exposure times.

In summary, in-camera shift technologies aren't easy to beat, if they can beat at all (IMHO).

Click to expand...

I think you overrate this new technology...

The main problem is that it behaves as if you had a higher resolution sensor but the resolution of the lens remains the same and this is in most cases the limiting factor. You can not make miracles except if you have a more than excellent lens !

Click to expand...

The inherent sensor resolution is exactly the same. Imagine you'd be using a monochrome version of it or a monochrome film simulation. That version already has the full "color" resolution, and the lenses are more than capable of supplying that resolution.

nixda said:

In quality, you beat if you use the technic of stitching multiple shots with a longer focal length. The final real resolution will be much higher.

I can make a real 100mp resolution image by stitching. No problems. But this is not possible with sensor shift, the real resolution will be much lower than the number of pixels.

Click to expand...

I often run into DoF issues when using longer focal lengths. Also, I do appreciate not having to get my 3D panorama setup out all the time and not having to stitch images. All that I think would be quite welcome in general.

But basically, you are arguing against any kind of higher-resolution sensor, because one can always stitch together images, even if the sensor had only 1MP.

nixda said:

I don't think this is revolutionary in fact, it has no effect on the optical quality. . This is just a nice to have.

Click to expand...

I don't know what you mean by "optical quality", but the image quality would be improved tremendously. Just compare the output of a Foveon sensor to that of a traditional sensor with the same resolution. It's quite easily visible.

Click to expand...

But Faveon sensors work a little different, in the sense that it has 3 different sensors al recording different color data, pixel shifting is basically panning the sensor around recording data from different angles (I think it's half a pixel) so basically is focus stacking by software and moving the pixels around, you can achieve the same results with focus stacking programs for about 1 or 2 hundred dollars and a little patience, now I'm not knocking it down I thing it's super cool and innovative I'm just saying its really not that useful everyday real life unless you already own a great tripod and shot mostly product photography, quite honestly I find the Faveon sensor much more impressive although it's slow and cumbersome to work with but you are right the files are stunning
--
http://instagram.com/victortrasvina/

VictorTrasvina said:

But Faveon sensors work a little different, in the sense that it has 3 different sensors al recording different color data, pixel shifting is basically panning the sensor around recording data from different angles (I think it's half a pixel) so basically is focus stacking by software and moving the pixels around, you can achieve the same results with focus stacking programs for about 1 or 2 hundred dollars and a little patience,

Click to expand...

There is the concept of super-resolution photography that captures the process precisely. And it's not trivial whatsoever. You can get an idea about it here.

VictorTrasvina said:

now I'm not knocking it down I thing it's super cool and innovative I'm just saying its really not that useful everyday real life unless you already own a great tripod and shot mostly product photography, quite honestly I find the Faveon sensor much more impressive although it's slow and cumbersome to work with but you are right the files are stunning

Click to expand...

With every camera, there are tradeoffs. The Foveon cameras are great at base ISO but fall apart pretty quickly at higher ISOs, and they are fixed-lens systems. Granted, the pixel-shift technology isn't for every situation, and I never claimed it is, but I personally would find it very interesting for my own type of photography. YMMV.

nixda said:

Christof21 said:

nixda said:

VictorTrasvina said:

nixda said:

What no one seems to point out is that you have been able to accomplish essentially the same thing, a high resolution image with multiple exposures with image stitching for years and years. Once a tripod comes into play nothing is convenient. And you are not limited by the programming in the camera. You want 100 MP? Fine.

Click to expand...

And don't forget it's useless on everything but shooting inert objects in studio (even landscapes are affected by wind) so you are right ! Might as well use focus stacking and get insane resolution and detail

Click to expand...

It is by no means straightforward to create high-resolution images from multiple shots. One way to achieve that is by using a longer focal length and stitching together images. But then, DoF would potentially be a problem.

One can also shift images and assemble higher-resolution versions from those. That is very difficult, though; there are no good tools as far as I know. And connecting the tools that cover certain aspects (3D alignment and averaging of images) is not a trivial task.

besides, I would contend that taking the images required for the mentioned techniques takes more time than a quick burst of three images that are automatically shifted by 1 pixel. Even for non-stationary objects, the method can be beneficial, depending of course on the amount of movement and exposure times.

In summary, in-camera shift technologies aren't easy to beat, if they can beat at all (IMHO).

Click to expand...

I think you overrate this new technology...

The main problem is that it behaves as if you had a higher resolution sensor but the resolution of the lens remains the same and this is in most cases the limiting factor. You can not make miracles except if you have a more than excellent lens !

Click to expand...

The inherent sensor resolution is exactly the same. Imagine you'd be using a monochrome version of it or a monochrome film simulation. That version already has the full "color" resolution, and the lenses are more than capable of supplying that resolution.

nixda said:

In quality, you beat if you use the technic of stitching multiple shots with a longer focal length. The final real resolution will be much higher.

I can make a real 100mp resolution image by stitching. No problems. But this is not possible with sensor shift, the real resolution will be much lower than the number of pixels.

Click to expand...

I often run into DoF issues when using longer focal lengths. Also, I do appreciate not having to get my 3D panorama setup out all the time and not having to stitch images. All that I think would be quite welcome in general.

But basically, you are arguing against any kind of higher-resolution sensor, because one can always stitch together images, even if the sensor had only 1MP.

nixda said:

I don't think this is revolutionary in fact, it has no effect on the optical quality. . This is just a nice to have.

Click to expand...

I don't know what you mean by "optical quality", but the image quality would be improved tremendously. Just compare the output of a Foveon sensor to that of a traditional sensor with the same resolution. It's quite easily visible.

Click to expand...

I was referring to the lens reeolution which is most often the limiting factor for the final image resolution.
Use a longer focal lens and you can improve a lot the resolution. In your original post, you were saying that the sensor shift technic could not be beaten, stitching can have much better REAL resolution in some cases even if there are disadvantages.