Olympus E-M5 II Image Samples for Multishot (Pixel Shift)

[capital]

In reference to Imaging Resource's samples, looking over the JPEG images of the multi-shot version (i.e. pixel shift), and comparing them to the Pentax 465D, it is remarkable how good this technology comes across. Apparently there is no color moire from this pixel shift tech, at least on the "Samuel Smith" green label  which the D800E, D645 and Z645 all glitch on. I am curious to see some real world examples, especially foliage on trees, as this pixel shift tech is most likely going to let down here unless there is a dead calm.

One other question, did Olympus do the right thing by up sampling pixel shifted outputs as opposed to providing a true color native resolution raw file? (At least that's what it appears that they doing).


http://www.imaging-resource.com/PRODS/olympus-e-m5-ii/olympus-e-m5-iiA.HTM

[AlterEgo]

One other question, did Olympus do the right thing by up sampling pixel shifted outputs as opposed to providing a true color native resolution raw file? (At least that's what it appears that they doing).

http://blog.lexa.ru/2015/02/05/pro_olympus_e_m5_2_i_ee_63_megapikselya.html

[davidgp]

Never saw how this shifting sensor technologies works, I'm assuming this only works with static subjects, right?

[SZRitter]

Never saw how this shifting sensor technologies works, I'm assuming this only works with static subjects, right?

If you watch The Camera Store's video, https://www.youtube.com/watch?v=0uQKYgQaNHQ, he shoots one of the city skyline. You can see motion blur in the heat exchangers, but it looks like Olympus did a fairly decent job of just making it look like motion blur, not weird artifacts.

[davidgp]

If you watch The Camera Store's video, https://www.youtube.com/watch?v=0uQKYgQaNHQ, he shoots one of the city skyline. You can see motion blur in the heat exchangers, but it looks like Olympus did a fairly decent job of just making it look like motion blur, not weird artifacts.

Many thanks for the link (I think I should subscribe to the camera store feed), I can not watch it now, but later today I will take a look...

[Guillermo Luijk]

Interesting those tests at imaging-resource comparing it to the sharpness (and moiré) queens: the Nikon D810 and Sony A7R.

All three are native camera JPEG, downsampled from 40Mpx to obtain a similar output size in the case of the EM-5. Although the Olympus keeps up as good as the best of the others in sharpness (even after the image has been resampled twice, one in camera and then by the imaging-resource guys), it's the only one that manages to avoid moiré. It seems clear that although it has no AA filter, oversampling with photosites that overlap by 1/2 pixel act as an excellent AA filter. Can't wait to see the RAW files:









Source: Olympus E-M5 II Review at imaging-resource.


Very important to find out if this mode can be used in applications that involve slight dynamism such as landscape, is to know how it behaves with moving subjects. A RAW file built from shifted captures could be prone to all kind of colour artifacts. From a test by Valentin Sama it seems that this mode handles very well subject motion, close to conventional motion blur:

Source: Olympus O-MD E-M5 MkII – Prueba de Campo.



100% JPEG crop


Regards

[Telecaster]

Seems to me the E-M5ii will make a killer digitizer for film and other relatively flat subject matter. I can see getting one specifically to re-archive my & my dad's transparencies & negatives. I'm impressed by the improvement in tonal gradation & reduction of aliasing artifacts in the 40mp examples I've seen so far. For general use I have no more interest in the multi-shot mode than in stitching or HDR stuff with existing cameras—which is to say some interest, not much—but I expect other folks will put it to good use.

Good to see a fully articulated rear screen too. A pity Olympus can't bring themselves to really beef up video capability…but maybe Panasonic has some say in that.

-Dave-

[Guillermo Luijk]

No words:

[Guillermo Luijk]

A test to fix artifacts in moving objects: since the artifacts have exactly 1 pixel width (they are like diagonal lines), downscaling to 50% and back to the original size through 200% upscaling provides a soft image artifact-free that can easily replace the affected areas using a layer mask:



I wonder if doing such corrections is needed with fluids like water, smoke, clouds,...  (soft shapes in uniform colours).

Regards

[AlterEgo]

Olympus Malaysia employee review = http://robinwong.blogspot.com/

[AlterEgo]

I'm impressed by the improvement in tonal gradation

Olympus EM5mkII it seems is using rolling electronic shutter in multishot mode, if it is the same panasonic sensor as in GH4/EM1 then it is 10bit raw = http://www.semicon.panasonic.co.jp/ds8/c3/IS00006AE.pdf

Electronic shutter (Full scan) = Full scan 10bit mode (22.5 fps) 1/22.5s~1/81300s (1/81300s step)

now firmware might present it differently, but

[capital]

Thank you Guillermo for showing this example, if there was a algorithm to detect this issue and minimize it, it would be better than trying to despot our own photos a myriad times over. Some photo editing is fun, this may become a bit tedious.

A test to fix artifacts in moving objects: since the artifacts have exactly 1 pixel width (they are like diagonal lines), downscaling to 50% and back to the original size through 200% upscaling provides a soft image artifact-free that can easily replace the affected areas using a layer mask:



I wonder if doing such corrections is needed with fluids like water, smoke, clouds,...  (soft shapes in uniform colours).

Regards

Imaging Resource also added a blurb about native 64 MP raw files accessible via a Photoshop plugin. Though there is still no mention of why it is better to not offer a true color 16 MP raw file, and allow people the option to upscale if they so choose. After reflecting on it further, since they have half pixel steps too, the file no longer is a native 16 MP dimension, instead it inserts those half pixels on each side of the pixels, so you get double the pixels on each dimension? i.e. 16 MP x 2^2 = 64 MP.

Actually, this thought gets me thinking, why don't they also have a mode that just does true color 16 MP? They could reduce the capture time down to just two x and y shift cycles

[Telecaster]

Olympus EM5mkII it seems is using rolling electronic shutter in multishot mode, if it is the same panasonic sensor as in GH4/EM1 then it is 10bit raw = http://www.semicon.panasonic.co.jp/ds8/c3/IS00006AE.pdf

What I was getting at was the lack of aliasing-induced tonal contamination in the E-M5ii samples. This isn't "gradation" in the sense of quantized dynamic range, but it's a disruption of tonality you can definitely see.

-Dave-

[BJL]

... Though there is still no mention of why it is better to not offer a true color 16 MP raw file, and allow people the option to upscale if they so choose.

It seem that a 16MP raw is also there, in a new ".ORI" file format (to have three color values per location?); the Olympus America press release says:
"When shooting in RAW+JPEG mode, the camera will save a 40M JPEG file, a 64M RAW (ORF) file and a 16M RAW (ORI) file. 64M RAW images may be processed using Photoshop CS4 and later with a required plug-in."

... why don't they also have a mode that just does true color 16 MP? They could reduce the capture time down to just two x and y shift cycles.

Yes, the might be an interesting option ... and could probably be done with a firmware update, if enough people ask for it.

[davidgp]

I think I will have to watch careful this camera, I was thinking to get something light for street photography (I tested the E-M1, lovely camera... ), but this could also work in some situations like a great landscape camera (in situations where you have no motion, I suppose waves in a beach will cause very strange patterns, or long exposures where you want grass moving or so on...)

[AFairley]

Seems to me the E-M5ii will make a killer digitizer for film and other relatively flat subject matter. I can see getting one specifically to re-archive my & my dad's transparencies & negatives. I'm impressed by the improvement in tonal gradation & reduction of aliasing artifacts in the 40mp examples I've seen so far. For general use I have no more interest in the multi-shot mode than in stitching or HDR stuff with existing cameras—which is to say some interest, not much—but I expect other folks will put it to good use.

Good to see a fully articulated rear screen too. A pity Olympus can't bring themselves to really beef up video capability…but maybe Panasonic has some say in that.

-Dave-

Plus the Oly 60mm macro is reputed to be wickedly sharp....

[BJL]

Can anyone comment on this naive idea for handling subject motion is some parts of the scene (leafs, clouds), when the main subject is still:
1) Detect regions of unacceptable subject motion by comparison between the eight frames.
2) Blend between two versions (a) the high res composite and (b) the 16MP first frame:
use the former where there are no subject motion problems, but blending towards the latter where needed.

Maybe this would have to be a manual fix, but for extra points, some slick software automation would be nice.

[SZRitter]

Can anyone comment on this naive idea for handling subject motion is some parts of the scene (leafs, clouds), when the main subject is still:
1) Detect regions of unacceptable subject motion by comparison between the eight frames.
2) Blend between two versions (a) the high res composite and (b) the 16MP first frame:
use the former where there are no subject motion problems, but blending towards the latter where needed.

Maybe this would have to be a manual fix, but for extra points, some slick software automation would be nice.

There was someone on a thread somewhere (maybe here?) who mentioned this is basically what they do with the Hasselblad MS backs. In essence, it has the same technical challenge.

[capital]

Can anyone comment on this naive idea for handling subject motion is some parts of the scene (leafs, clouds), when the main subject is still:
1) Detect regions of unacceptable subject motion by comparison between the eight frames.
2) Blend between two versions (a) the high res composite and (b) the 16MP first frame:
use the former where there are no subject motion problems, but blending towards the latter where needed.

Maybe this would have to be a manual fix, but for extra points, some slick software automation would be nice.

BJL, It is a good idea, I could see its implementation, though the reality is if only one consumer camera uses it, a la Foveon, then the likelihood of a commercial product that addresses this computational photography homework problem diminishes. As an aside, Guillermo mentioned above about a quick method: down scale, upscale and then blend as needed.

It would be of interest if Olympus will allow the export of any of the 8 individual 16 MP subsamples as a raw file.

[Guillermo Luijk]

Can anyone comment on this naive idea for handling subject motion is some parts of the scene (leafs, clouds), when the main subject is still:
1) Detect regions of unacceptable subject motion by comparison between the eight frames.
2) Blend between two versions (a) the high res composite and (b) the 16MP first frame:
use the former where there are no subject motion problems, but blending towards the latter where needed.

Maybe this would have to be a manual fix, but for extra points, some slick software automation would be nice.

It's exactly what I had in mind when I did the manual correction over Dpreview's cars: detect moving areas by frame to frame comparision, and build the output image/RAW file differently for the two cases. But this should be fixed in firmware since there is no access to the eight single captures.

Regards