[biz-engineer]

Everyone will think Sony invented it.

Yes, that's what I understand if I read comment about the A7RIII.

---------- Post added 26-10-17 at 08:44 ----------

Actually Sony tries a crippled version of it:

Yeah, the Pentax pixel shift is different.

---------- Post added 26-10-17 at 08:57 ----------

Sony product life cycles seems to be short lived. They release something like 1 new full frame camera every other year with a few hardware tweaks and unfinished firmware and then publish frequent firmware updates to keep Sony being told about. That's a new way of marketing camera, that also Fuji embraced, fairly different from the traditional Canikon Pentax way.

[beholder3]

As for the delay. .5 second (minimum) delay is unfortunate but it's not the end of the world.

Just as 1 MPx sensor is not the end of the world.

You have this delay three times, so you need have a subject utterly motion free for 2 seconds. That is 100% worse than Pentax. And worse here means 100% more motion artifacting.

[ElrondElensar]

As die hard fan of pixel shift, taking 4 photos and merging them later is NOT the same as true pixel shift.

You can get the same result on any camera by shooting in burst mode and combining the images in photoshop. Here is how to do it

A Practical Guide to Creating Superresolution Photos with Photoshop

[Rondec]

You can get the same result on any camera by shooting in burst mode and combining the images in photoshop. Here is how to do it

A Practical Guide to Creating Superresolution Photos with Photoshop

You do understand that what the end result of both the Pentax and Sony pixel shifts is not a "super-resolution" file, but a file with less noise and better color depth? As I mentioned in another thread, if your goal is super resolution you will probably have a lot better results shooting panoramas with a longer lens. But the point here is to give a "best possible quality" base file with no moire that you can then process. But it is still only 36 megapixels (or in the case of the A7r III 42 megapixels).

I just wonder how useful the Sony implementation since most scenes move over a 2 second period of time. If you have much movement you are better off just using a single file -- at least for Pentax.

[ElrondElensar]

You do understand that what the end result of both the Pentax and Sony pixel shifts is not a "super-resolution" file, but a file with less noise and better color depth? .

That's also true for Olympus and Hasselblad if you downsample the image to original size. All three do exactly same thing.

Download the Olympus high-resolution image from dpreview, and downsample it original size. You will less noise, more DR, less moire, and more resolution, than the single shot image, which is to be expected, identical benefits to Pentax or sony or if you do the same thing in photoshop with any camera as the link I provided shows you how to do it handheld with any camera.

[normhead]

Interesting Sony saw fit to join the party, albeit a bit late. Given the number of forum members who bought Sony's to use their Pentax lenses on FF, something which has completely dried up since the K-1 was introduced, I wonder if this is a strategy to try and keep that income stream from dying. It doesn't seem like enough, but who knows?

[ffking]

It's been what, a year since the last version and this one blows it out of the water.

I'm sure they have many teams working on different aspects separately - that's certainly borne out by the appearence

[serothis]

You mean moire or moiree reduction artefacts?

I meant to post this yesterday but this is the anomaly

OOC JPEG


Processed RAW

Just as 1 MPx sensor is not the end of the world.

You have this delay three times, so you need have a subject utterly motion free for 2 seconds. That is 100% worse than Pentax. And worse here means 100% more motion artifacting.

That's a tad melodramatic. Pentax's implementation is quick but it's by no means instantaneous. I would more liken it to a 36MP vs 16MP camera. One is clearly better than the other but the latter is still capable.

You can get the same result on any camera by shooting in burst mode and combining the images in photoshop. Here is how to do it

A Practical Guide to Creating Superresolution Photos with Photoshop

I read through the article and while it's an interesting approach that might achieve similar results in certain situations, I stand by my claim that it is not the equivalent of pixel shift. Here's why:

• The method outlined in the article requires ~20 images which takes time to shoot and lots of storage space. It then takes time to merge. Pixel shift creates 1 file (equal in size of 4 files) and also takes time to merge but I bet that it is substantially less time than aligning it in photoshop.
• The method calls for random moment and image alignment vs Pixel shift which uses precision movements (more importantly known movements) and specialized algorithms to make accurate calculations.
• The article recommends hand held at shutter speeds 1/2*focal length. Specifically that the natural hand movement acts at the "pixel shift". That means you can't do any tripod work. So images like the below are impossible. It also suggests that higher shutter speeds plus burst mode would be less effective.
  
  This was 20s exposure.
  

[beholder3]

One is clearly better than the other but the latter is still capable.

Sony's failed cripple version is multiple times slower and lacks half of the development options.
That is like 3 leagues poorer. Not even close to capable.

[Rondec]

I meant to post this yesterday but this is the anomaly

OOC JPEG


Processed RAW




That's a tad melodramatic. Pentax's implementation is quick but it's by no means instantaneous. I would more liken it to a 36MP vs 16MP camera. One is clearly better than the other but the latter is still capable.



I read through the article and while it's an interesting approach that might achieve similar results in certain situations, I stand by my claim that it is not the equivalent of pixel shift. Here's why:

• The method outlined in the article requires ~20 images which takes time to shoot and lots of storage space. It then takes time to merge. Pixel shift creates 1 file (equal in size of 4 files) and also takes time to merge but I bet that it is substantially less time than aligning it in photoshop.
• The method calls for random moment and image alignment vs Pixel shift which uses precision movements (more importantly known movements) and specialized algorithms to make accurate calculations.
• The article recommends hand held at shutter speeds 1/2*focal length. Specifically that the natural hand movement acts at the "pixel shift". That means you can't do any tripod work. So images like the below are impossible. It also suggests that higher shutter speeds plus burst mode would be less effective.
  
  This was 20s exposure.
  

I don't know if you looked at the crops of this "super-resolution," but there is a tiny improvement in detail, but nothing like you would expect to see when combining 20 files. If I shot twenty images with a little longer lens and stitched, I would be able to zoom in three or four times as much and see good detail. All you are seeing here is a slight bump in detail and slight reduction in moire, but nothing dramatic. Combining four images in pixel shift certainly seems to give better results, in my experience, as long as it is a still day.

[serothis]

Sony's failed cripple version is multiple times slower and lacks half of the development options.
That is like 3 leagues poorer. Not even close to capable.

Again I think you're being overly dramatic. How many PP development options does pentax have?

In-camera - I haven't had much luck with this
PDCU - I'm not a huge fan of.
RawTherapy - I use religiously.

There's probably others I don't know about but it's not like Pentax has an overpowering mountain of options. And I have no doubt that in time 3rd party software, like rawtherapy, will also support Sony.

And if you look at my river picture above, that's a 20 sec exposure. An extra 2 seconds isn't going to make a huge difference when the total exposure time is 1min 20sec.

Again, I think Pentax's implementation is probably better but Sony's is still (on paper) useful. Just a few more limitations.

-----

On a side note there might be a very tangible benefit to the delay. As I read the press material they stated that the delay was .5, 1 and 2 seconds. Although this is just a guess, This suggests that you can change the delay duration. It hasn't been mentioned yet but IF sony allows a flash to be used while in pixel shift mode, that means you can delay each shot long enough for your strobes to recycle. That's something that pentax's implementation can't do (and I wish they would). This would be a huge boon for product photography.

[Rondec]

Again I think you're being overly dramatic. How many PP development options does pentax have?

In-camera - I haven't had much luck with this
PDCU - I'm not a huge fan of.
RawTherapy - I use religiously.

There's probably others I don't know about but it's not like Pentax has an overpowering mountain of options. And I have no doubt that in time 3rd party software, like rawtherapy, will also support Sony.

And if you look at my river picture above, that's a 20 sec exposure. An extra 2 seconds isn't going to make a huge difference when the total exposure time is 1min 20sec.

Again, I think Pentax's implementation is probably better but Sony's is still (on paper) useful. Just a few more limitations.

-----

On a side note there might be a very tangible benefit to the delay. As I read the press material they stated that the delay was .5, 1 and 2 seconds. Although this is just a guess, This suggests that you can change the delay duration. It hasn't been mentioned yet but IF sony allows a flash to be used while in pixel shift mode, that means you can delay each shot long enough for your strobes to recycle. That's something that pentax's implementation can't do (and I wish they would). This would be a huge boon for product photography.

Pixel Shift probably adds very little to any image if any individual exposure is more than 0.5 seconds in length. It is an interesting exercise to use Raw Therapee to extract and individual exposure and compare it to the pixel shift exposure. What you find is that in many situations there is little benefit. I mainly see benefit in situations where shutter speed isn't too long, there is no movement, and the dynamic range is really high. Otherwise I don't think it is worth it (and I really like pixel shift).

[biz-engineer]

The article recommends hand held at shutter speeds 1/2*focal length. Specifically that the natural hand movement acts at the "pixel shift". That means you can't do any tripod work. So images like the below are impossible. It also suggests that higher shutter speeds plus burst mode would be less effective.

This is a method of spatial oversampling. The same method is use in high speed oscilloscopes where the same signal is sampled by multiple channels each having a sampling clock is phase shifted in small increments. So, for example, if the same signal would be sampled by 4 channels, each channel with 90 degree phase shift, the time resolution is 4 times faster than the time base of the scope.

How does that related to super resolution with multiple shots aligned? It is the same technique as spatial over sampling used by high speed scopes. The difference being that the phase shift is random, but each photo should be equally sharp. So , shooting at 1/2*FL would add some blur and it's not good at all. However, shooting on tripod is not a problem is the camera position is moved for every shot. Also, even 20 shots is not sufficient to ensure evenly spread camera position, 30 to 50 shots would be better to achieve super resolution on the basis of random camera position.

Between the constrain of very stable tripod for a sequence of 4 pixel shifted images and shooting a few dozens of shoots, I don't know what method is easier, IMO both method aren't as straightforward as a single shot.

Anyway, as I wrote in anther thread, unless printing XXXL, for the rendering of an image, more resolution is pretty useless compared to the use of a larger sensor, however, more resolution makes camera slower to operate while lower pixel count on a larger sensor brings better images at fast speeds.

[JohnBee]

the details and video can be found here: Sony a7r III mirrorless camera officially announced | Photo Rumors

The a7r III brings numerous improvements over the now older a7r II
With that said, I think we can cut Sony some slack on their first attempt at pixel shift technology. Granted, it won't likely be anything more than a marketing novelty, but it does however, show that Sony is paying attention to what's going on outside of their own product lines.

My guess is that they'll either continue to refine this feature along with their mirrorless bodies or give up on it altogether in favor of higher sensor performances in the future.

PS. having played with pixel shift, I can't bring myself to justify it's use as anything other than a novelty in any camera at this current time. - and I'm a huge fan

[bclaff]

That's also true for Olympus and Hasselblad if you downsample the image to original size. All three do exactly same thing.

Download the Olympus high-resolution image from dpreview, and downsample it original size. You will less noise, more DR, ..., than the single shot image, which is to be expected, identical benefits to Pentax or Sony ...

You can clearly see the effect for Olympus by looking at Photographic Dynamic Range (PDR) - Olympus High Res Shot Mode Special Edition
Perhaps in the future I'll have appropriate measurements for a Pentax Pixel Shift Resolution Special Edition as well.