[konraDarnok]

I think you are missing the point here, Astrophotographers typically do very, very long exposures on a moving platform to compensate for the rotation of the earth, with their lenses trying to resolve objects through 100Km of atmospheric gases.

Astrophotography is considerably more complicated than what most photographers do during the daylight hours.



ohh sure who needs hardware when software can solve everything.

/sarcasm.

No, I'm not missing the point. I've done some astrophotographic stitching. The exposures are not "very long." In fact, many people just record video and then stack thousands of frames of very short duration. This doesn't even require a motorized equatorial mount that maintains the proper orientation. Altazimuth mount will work as well.

Your "sarcasm" isn't really warranted.

[mecrox]

I don't think THAT much processing power or software is needed at all. I imagine that the sensor would move in the same preset direction each time. So software doesn't need to find corresponding points and warp and stitch those together. It just needs to combine multiple images the same way each time.

Yes, a tripod would be mandatory? So? Aren't our cameras married to tripods anyway? This would finally be a reason to use SR!

Why would a tripod be mandatory if the camera took one image but shifted a copy or two in software instead of taking a second image, then used some kind of average system? It's HDR-like but not as we know it, Jim.

[Clavius]

Why would a tripod be mandatory if the camera took one image but shifted a copy or two in software instead of taking a second image, then used some kind of average system? It's HDR-like but not as we know it, Jim.

The rumor says: "Pentax K3 with 40 MP via sensor shift". Not: "Pentax K3 with 40 MP via software shift". Please see the opening post.

[Na Horuk]

Why would a tripod be mandatory if the camera took one image but shifted a copy or two in software instead of taking a second image

The idea is that it shifts the actual sensor so that it captures the detail that otherwise gets lose "between" the pixels. Software can only guess what that info should be, which is passable at best and downright ruining photos at worst. Bayer sensors already do some guesswork.
It would essentially take a couple photos and merge them. And it would be interesting if it could also use a different ISO for each part and increase dynamic range at the same time.

[Caat]

This doesn't strike me as a particularly marketable feature for most people.

I might use it as I do quite a lot of photographic cataloguing for a museum and imaging of objects for research applications. However outside this, product photography and some landscape work I don't see it being very appealing - given the need for a stationary subject. Landscapes would depended upon how much moveable plant/animal life there was and what the delay between the frames is.

[mecrox]

The idea is that it shifts the actual sensor so that it captures the detail that otherwise gets lose "between" the pixels. Software can only guess what that info should be, which is passable at best and downright ruining photos at worst. Bayer sensors already do some guesswork.
It would essentially take a couple photos and merge them. And it would be interesting if it could also use a different ISO for each part and increase dynamic range at the same time.

Ah, I see now. You're the first person who's explained what is going on. Thanks. Different ISO for each part sound a very interesting idea.

[mamethot]

Has anyone here ever saw a raw file from a phone camera?

I have not. I searched the matter though and what I was able to find are mentions that the raw file isn't good at all and that your phone camera is only usable because of software correction.

Now I'm not saying that software is the answer to anything but if software can make a multi-million dollars F-35 (which would be near impossible to pilot) or a Boeing C-17 as easy to fly as a Cessna 152, then it's possible that the image out of your camera might be improved by using software as well. Of course it's not at the same level of costs, hardware and targeted audience, but scaled to the appropriate level, it can help.

Actually the software enhancements are already present in most camera and there will be more and more in the future.

As for marketability (?) I think that if IQ is better, it will be easy enough to market.

[Clavius]

This doesn't strike me as a particularly marketable feature for most people.

I might use it as I do quite a lot of photographic cataloguing for a museum and imaging of objects for research applications. However outside this, product photography and some landscape work I don't see it being very appealing - given the need for a stationary subject. Landscapes would depended upon how much moveable plant/animal life there was and what the delay between the frames is.

It completely depends on what the delay between frames is. And that's why we cannot speculate yet on what areas of photography it will be usable in.

Moreover, there's more features on our cameras that not everybody uses all the time. It's very usefull tool for the foolbox, imho.

[Caat]

It completely depends on what the delay between frames is. And that's why we cannot speculate yet on what areas of photography it will be usable in.

Moreover, there's more features on our cameras that not everybody uses all the time. It's very usefull tool for the foolbox, imho.

True, but there's a development cost associated with it. If it's low that's fine.

[Clavius]

True, but there's a development cost associated with it. If it's low that's fine.

Well, the required hardware (SR) is already there. And if it gives an edge over the other brands that extra developement is more then worth the effort.

[Fogel70]

I don't think SR can be used for this type of shift as it would be too difficult to make it with the precision and repeatability needed for this.
So Pentax probably has to replace the SR system with piezo shift mechanism like on previous cameras with this function.
SR is designed to move the sensor in a controlled speed and controlled direction, not to position the sensor within a few hundreds of a micrometer.

Not sure if many user want to replace shake reduction with this on a APS-C body, but on 645D it might be a killer.

[MouLaG Ôfr]

I think it is possible and it even exists for now : see the astrotracer function with the O-GPS1

It works with very small and repeated sensor shifts, for what I understand ;-)

GPS unit O-GPS1 - RICOH IMAGING UK LTD." style="padding-top: 3px; padding-left: 1px; padding-bottom: 1px;">ASTROTRACER for effortless astronomical photography


When mounted on the PENTAX K-5 or K-r camera body, the O-GPS1 also offers the advanced ASTROTRACER function,** which couples the unit with the camera’s SR (Shake Reduction) system for the effortless tracking and photographing of celestial bodies. The unit calculates the movement of stars, planets, and other bodies using the latitude obtained from GPS data and the camera’s alignment data (horizontal and vertical inclinations and aspect) obtained from its magnetic and acceleration sensors, then shifts the camera’s image sensor in synchronization with the movement of the objects.*** As the result, stars and other bodies are captured as solid points rather than blurry streaks, even during extended exposures. It also makes astronomical photography much simpler, as it requires only a tripod and eliminates the need for an additional accessory such as an equatorial telescope.
** This function is available only when the O-GPS1 is mounted on a PENTAX digital SLR camera body equipped with a magnet-driven SR system.
*** The duration of ASTROTRACER operation may vary depending on photographic conditions.

[monochrome]

RE: Development cost, think in terms of releasing a technology most useful for the 645DII on a high volume dSLR to spread out the costs, as was done with the K-7 and the 645D.

[Fogel70]

I think it is possible and it even exists for now : see the astrotracer function with the O-GPS1

It works with very small and repeated sensor shifts, for what I understand ;-)

You don't need the same precision for the astrotracer function. For astrotracer it might be good enough with resolution of 1-2 pixels as it will be constant moving over a long period of time. I don't think I seen any picture taken with O-GPS1 that are pixel perfect when viewing at 100%, but they are still pretty good.

For adding resolution you probably need close to 1/10 of a pixel resolution, as otherwise the error will be to large and the added data will be garbage, so you might get better result by upsampling the image in PP.

[mgbirder]

The idea is that it shifts the actual sensor so that it captures the detail that otherwise gets lose "between" the pixels. Software can only guess what that info should be, which is passable at best and downright ruining photos at worst. Bayer sensors already do some guesswork.
It would essentially take a couple photos and merge them. And it would be interesting if it could also use a different ISO for each part and increase dynamic range at the same time.

I’m not sure if it is the same principle, but the Fuji X-Trans sensor does just that. I have the Fuji x10 and along with its DR Settings (DR100, DR200, DR400), it has a nifty low light feature, it takes 4 images at different ISO settings and stitches them together in camera, obtaining an image with higher DR and lower ISO Noise. The downside is that the resolution is spitted from 12 to 6 MP, and not multiplied as the rumor of the Pentax K3 suggests.