Olympus OM-D Em5 Mark II: High resolution shot mode

[jojm]

Olympus OM-D Em5 Mark II, the last launched of the Olympus MFT expert cameras, has many useful utilities for a micro-minerals photographer like me.
First of all, there is a perfect silent mode which allows you to shot without any vibration at all, first and rear curtains shutters are totally electronic in this mode. Some of the Lumix/Panasonic camera have also this feature but with a limitation about exposition which cannot be longer than 1s, not enough for me which use very soft lightning and low ISO shooting.
With the Wifi connection and the Olympus Image Share software, you are able to control your camera from your tablet or your smart-phone.
Without the use of silent mode but OK with anti-vibration mode (Electronic First Curtain Shutter), you could also control your camera directly from a PC via the software Olympus Capture.

The third mode for shooting which could interest micro-minerals photography is the High resolution mode.
With this mode, the camera takes 8 pictures of the same scene moving the sensor by 0.5 px between each shot and complies the eight images into one high resolution one.
http://www.imaging-resource.com/PRODS/o ... iTECH2.HTM
This mode needs a perfect immobility of the subject but what is more quiet than our minerals under our lenses !
If your exposure time is 2s, you will need to wait about 25 s to get an HR image.

By following this mindat.org link, you could zoom on the high format picture by scrolling your mouse wheel on a PC or with your fingers on tablets and smart-phones.

The original picture was more than 40MP large but with the ratio 3:2 used for cropping, it is now 35MP (7296x4264).
For this shot session, I've used a lens made for the old Nikon polarizing microscope POH-2 when it was use with the Universal stage. It is the Nikon U x10 NA=0.22 WD=16,4 mm. Technical data sheet speaking, it is not the best I own but I like its bokeh and the room given for lighting by the long WD.
The shot session was quite long, nearly one hour for 86 images.
For lighting, I've used 2 IKEA Jansjo LED with diffusion around the subject.

[rjlittlefield]

Jean-Marc, thank you for the posting. Not only did I learn some technical info, but the image is also great -- very nicely framed, illuminated, and processed. The image is well worth the hour to shoot it!

About the technical info, I was not aware that the OM-D Em5 Mark II has such nice shutter features.

One thing I am curious about: how did you step focus and trigger the camera?

The reason I ask is that Olympus cameras are famous for having a "multifunction" connector in which USB and wired shutter release use the same physical connector. As a result, when you plug in USB for wired tethering, you're no longer able to use wired shutter release from a StackShot controller.

Does this camera not have that problem, or does going wireless solve this problem, or did you do something else?

--Rik

[jojm]

Hi Rik,
I did step focus with the micrometer of an old microscope body with mini graduation each 0.002 mm. For this stack, I used a 5 graduations step so 0,01 mm between each image.
I triggered using an Ipad mini connected by Wifi.
I do not have a stack-shot system so it will be difficult for me to answer your question but I can confirm that there is only one port for USB and remote control connecting.

[rjlittlefield]

Great, thanks for the info.

--Rik

[Chris S.]

Jean-Marc, formidable!

I echo Rik’s appreciation of both your image and technical information. Your proustite is one of my favorite micro-mineral shots, ever. I’ve returned to it several times, zooming in and out looking at details. The lighting shows the surface structure superbly. The bokeh, as you remarked your lens does well, is wonderful.

. . .High resolution mode.
With this mode the camera takes 8 pictures of the same scene moving the sensor by 0.5 px between each shot and complies the eight images into one high resolution one.

The tests of this feature, in your link, were convincing.

So I have a pair of questions for Rik. Rik, suppose one wanted to do something similar with a camera body having a larger sensor than m4/3, and wished to take a do-it-yourself approach. The hardware end seems simple enough, at least at the conceptual level. A pair of piezoelectric motors would have the resolution to move an ordinary camera body repeatably by half a pixel, and a controller would be simple enough to build. But the software end is not so obvious to me—once one obtained the eight pictures, how would one combine them? Would Zerene Stacker PMax do it?

And I’m guessing that such an approach would only provide improvement for those lenses that out-resolve a given sensor. Seem right?

--Chris

[ChrisR]

Interesting.
I too had put that particular lens into the "not especially good" category.
I'm mulling over the differences between using small sensor movements to build up more data, and using a higher magnification lens with stitching of the images.

[rjlittlefield]

So I have a pair of questions for Rik. Rik, suppose one wanted to do something similar with a camera body having a larger sensor than m4/3, and wished to take a do-it-yourself approach. The hardware end seems simple enough, at least at the conceptual level. A pair of piezoelectric motors would have the resolution to move an ordinary camera body repeatably by half a pixel, and a controller would be simple enough to build. But the software end is not so obvious to me—once one obtained the eight pictures, how would one combine them? Would Zerene Stacker PMax do it?

The software end is also conceptually straightforward, but it's different from PMax.

What you would need is a simple image merging program that would input four images of size WxH and output a single image of size 2Wx2H, generated by pixel interleaving.

Assuming zero subject movement and perfectly equal exposures among the 4 input images, the resulting output image would be equivalent to a single image captured by a 2Wx2H sensor with a certain low-pass optical filter in front of it.

No doubt you'd then want to back out the effect of the filtering, but I think the nature of that is just the same as what we already have to do to compensate for diffraction blurring and anti-aliasing filters.

And I’m guessing that such an approach would only provide improvement for those lenses that out-resolve a given sensor. Seem right?

Pretty much true. As always, you gain something by capturing more photons at a finer spatial resolution. But unless the optical image actually out-resolves the sensor, the gains are mostly in noise reduction.

On the other hand, at low magnification covering a large field, it's simple to demonstrate the need for a gazillion pixels, as we've discussed before (for example HERE and at other references within that same thread). Covering only a 2 mm field (10X on MFT), maybe not so simple.

I'm mulling over the differences between using small sensor movements to build up more data, and using a higher magnification lens with stitching of the images.

Add to the list, use a larger format camera or a scanning back, with appropriate optical relays to fill the larger "sensor".

The methods are not mutually exclusive. I have visions of somebody equipping a GIGAmacro system with a Hasselblad H5D-200c MS camera (50 megapixel sensor, expanded to 200 megapixels by sensor-shifting).

--Rik

[ChrisR]

Sure - my"list"only extended to do-it-yourself apporoaches. It would be possible as Chris S describes, to make the camera translations of a couple of microns in two directions with hardware "we" have around. I'm pondering how that compares with stitching at a higher mag. Both have pros and cons.

[TheLostVertex]

The methods are not mutually exclusive. I have visions of somebody equipping a GIGAmacro system with a Hasselblad H5D-200c MS camera (50 megapixel sensor, expanded to 200 megapixels by sensor-shifting).

And what are your visions of that persons hair after they have finished stacking and stitching all the 200mp images together?

I can imagine it would be pretty simple to produce very large and difficult to handle images like this. Difficult to handle since stacking 200mp image would be pretty cumbersome to start with, then the final output image from stitching would be quite a beast. I do not even want to think about doing it beyond 8bit.

[rjlittlefield]

Hair? What hair??

As for stacking 200 megapixels, that's not common but it's definitely routine practice for at least one Zerene user. That's how I happen to know about the Hasselblad camera.

Besides, it's not really such a big number. Storage and processing time is linear in the pixel count, so each of those 200 megapixel images is equivalent to only about 6 images from a Nikon D800/810, or 4 images from the new (soon to be released) 50 megapixel Canon 5DS. You just need a 32 GB machine to do 200 megapixels comfortably.

--Rik

[Charles Krebs]

Jean-Marc,

Wonderful image! The "zoom-in" capability is a marvelous usage of this technology.

And the start of an interesting conversation...

Congratulations to Olympus for incorporating this into a reasonably priced "consumer" camera.

This approach has been used in some rather expensive "scientific" cameras for a while. The one that comes to mind is the Zeiss AxioCamHR. It uses a small 1.4Mp sensor (1388x1040) together with "pixel shifting" to produce a 13Mp (4154x3120) image. There are some interesting diagrams and discussions on pages 6 and 7 of this brochure. Interesting how Zeiss emphasizes the potential improvement in color representation (with a Bayer filtered sensor) in addition to the increased resolution:
http://www.krebsmicro.com/pdf/AxioCamHR.pdf

[Alex H]

But the software end is not so obvious to me—once one obtained the eight pictures, how would one combine them?
--Chris

PhotoAcute (http://www.photoacute.com/) might do the trick. I tried it once long time ago, but not recently and not with this type of images.

[Asha]

The Olympus "Pro" line of mft cameras is really awesome--I've been drooling over those cameras for a while now. Thanks for sharing your insights and your resulting image!

[Pau]

This approach has been used in some rather expensive "scientific" cameras for a while. The one that comes to mind is the Zeiss AxioCamHR. It uses a small 1.4Mp sensor (1388x1040) together with "pixel shifting" to produce a 13Mp (4154x3120) image. There are some interesting diagrams and discussions on pages 6 and 7 of this brochure. Interesting how Zeiss emphasizes the potential improvement in color representation (with a Bayer filtered sensor) in addition to the increased resolution:
http://www.krebsmicro.com/pdf/AxioCamHR.pdf

Pentax has implemented this last approach in its new K3 II
http://www.dpreview.com/articles/723818 ... ii?slide=3

[lothman]

Pentax has implemented this last approach in its new K3 II
http://www.dpreview.com/articles/723818 ... ii?slide=3

and the System of Pentax does quite well, comparison Pentax K3ii, Nikon D810, Olympus E-M5 II at Imaging Resource