Sunday, 28 December 2014

Sensor shift in the New Olympus EM5 Mk.2.

The refresh of this v. successful model by Olympus will be introduced in February, so all I have is some rumors. Movie rate, EVF, many of the features introduced by the latest Olympus models will be there, but the killer feature will be sensor shift (SS).

So far we know v. little: that it takes 8 pictures byf the sensor by 1/2 pixel, and achieves an image of 40 Mpx. We also know that IBIS stabilisation works in a bracket of 5 pixels . Since  1/2x8 = 4 pixels, so my bet is that sensor shift will happen as part of IBIS, with no need of a tripod.

By my early experience with a superresolution program called 'PhotoAcute' I know that you must slightly shift an image after the first so the program can compare pixels, and increase the information as a result. The program, created by a Russian mathematician,  was v. processing intensive, since each pixel of the preceding picture had to be compared with each of the following one.

Thankfully camera processing power doubles each year, so presumably each 8 pictures can be compared in a reasonable time, seconds instead of minutes.

We also know that the 'old' E-M5 could shoot at 10 fps, so again 8 frames can well be taken  within just one second.

The old E-M5 'Its going to be redesigned anyway.

I assume that stotal shutterspeed of the 8 frames shouldn' be more than one second: shutterspeed defines how fast the objects in front of you can move, fluttering leaves in a windy day being the typical problem for the landscapist. The other problem is people moving through the landscape, but freezing that  should be well within the capacity of the new system, provided one operates in fair weather.

What is the rationale for SS? First you don't need to change system anymore to have v. high definition. You don't even need to change lenses with more resolving ones, since the process is sequential.

At any rate the feature is excellent news for landscapists, portrait and macro users, down to PJ. An Editor is always happy to crop pictures (and a noob too).

People are asking if there will be less noise and better colours. Going by PhotoAcute, I'd say that the colours stay the same, but noise is filtered, while the sofware compares pictures.
A user also advocated more Dynamic Range a' La HDR.
But I'd rather bet that exposure is determined by the first shot and doesn't change for the following 7. The opposite would be a loss of time. So HDR will stay a separate feature, like it was in Photoacute.

BTW the E-M5 already has 12 bits of DR, so it's more than enough. Introducing more would mean to lessen the contrast and having flat pictures. Same goes for colour, it probably stays the same of the first picture, so what you earn with SS is really more detail and less noise.

All this was based on a pure deduction a la Sherlock Holmes, (!) but it might be v. different of what Olympus' engineers have concocted. Then wait for February and check how much of this set of assumptions is wrong :)

To me staying with one system only, and only one set of lenses is a remarkable advantage, and SS has stopped me in my tracks while taking an interest in Sony's A7 and having a fit of GAS. So well done, Olympus, with your 40 Mpx! My wallet is grateful...

When you are able to dabble with single pixels, small sensors still have the advantage. Less processing to do, quicker reactions of the camera.
Olympus' genius is to have designed a superresolution device working within IBIS, and therefore avoiding the need for a tripod like the Sony A7r is reported to need.

Now to recap: Sensor shift must work within IBIS.  IBIS shift works within 5 pixels. If so 1/2 px x 8 frames = 4 px, well within IBIS.

EM5 also has a fast 10 fps so all could happen within a second. With a new double speed processor integrating the eight shots shouldn't take more than a handful of seconds, like an Art Filter.The result is more than twice the resolution and less noise.

No comments:

Post a Comment