Questions about Olympus "50MB" and "80MP" images

Started 2 months ago | Discussions thread
Jan Chelminski
Jan Chelminski Senior Member • Posts: 1,934
Re: Questions about Olympus "50MB" and "80MP" images

Astrotripper wrote:

mount evans wrote:

I have an OM-D E-M5 mk I (actually, it wasn't a mk anything when I got it) and won't be upgrading anytime soon, so this is a bit like a monk asking questions about sex.

I was reading about the newer versions of the OM-D E-M5 and M1 taking higher resolution images by moving the sensor around and combining multiple images. One of these modes moves the sensor around in a square one pixel spacing on a side and combines the raw data accordingly so that you have Red, Green, and Blue channel information for each pixel (except for the ones at the edges, and redundant Green channel information for the rest).

Olympus does not have such a mode you describe. HR (the tripod variant) always takes 8 shots. Two sets of fours pixel shifted shots, offset by half a pixel.

This image should have the same resolution (plus one row and one column) as an ordinary image, with more color information, shouldn't it? Why, then, is this image referred to as a 50MP image?

50MP is the more reasonable estimation of the amount of detail in the image. So that's what the camera generates. This is about what you would also export to when processing the raw file, unless you like really large files for no good reason. I usually export to around 40mp.

Supposedly the "80MP" image then moves over 1/2 of a diagonal spacing and does this all over again, then fakes up a double-resolution image.

It's called supersampling.

I would think that there would be a better pattern for doing this, but I'm not on the design team.

Well, you can do random movements, but then you need more input data. Olympus does exactly that in Handheld HR mode.

I can sort of understand why these are called "80MP," though it is not entirely honest.

How is it not honest? It literally produces an 80mp raw file.

Another point that I'm sure no one cares about, as I do the math for the Raleigh criterion for resolution, the 20MP OM-D sensor is diffraction limited for 530nm green light at f/2.6 in 50MP mode--unless you have a larger aperture than that, you shouldn't be able to get any better spatial resolution by moving the sensor over by a half diagonal.

You are still using a 20mp sensor to capture the image. Supersampling doesn't really change anything about that. The side effect of this is that it also acts as a strong antialiasing filter. This is why the raw image is so soft when viewed at 100% and requires different processing. And why it doesn't really make sense to export to 80mp.

You might get some noise reduction by combining multiple exposures,

Not might. You do get it. Always.

but you could have done the same by using a longer exposure at a lower ISO.

There is no ISO 50 on Olympus cameras.

It would take a lens f/1.8 or faster to "do justice" to the extra spatial sampling of this mode.

Not really. While diffraction hits HR mode images quicker than normal images, it's still around f/5.6. Basically, on a normal image you cannot really perceive a drop in sharpness from the peak until you hit somewhere like f/8 (for super sharp lenses that beak before f/4) or f/11 (everything else). On HR image, you can actually see it far better and sooner.

Has anyone confirmed that this trick actually results in sharper images?

Sharper? In the strict sense of that work, no. At 100%, the image is actually a lot softer because of the AA filter effect I mentioned above.

Anyway, why don't you just play with it yourself? Get the samples from DPR and/or Imaging Resource and open them in whatever raw developer you use. Fun fact, unlike many other high res implementations, Olympus HR raw files are just regular raw files, pretty much all software supports that as there is no extra processing required on the application's side.

What f-number were they using?

In practice, you want to use your lens at peak sharpness. HR mode is limited to f/8 and that is for a reason. You can easily see softening going down from f/5.6 to f/8, with any good lens. HR image will also show weakness in lenses that were hard to see on regular shots. Does it have a slight field curvature that only goes away at f/4? Well, with HR shot you might see that it actually does not go away at f/4

(editing snafu)

Good points, but I do also think it is important to question statements made (and subsequently repeated by others) by Olympus ('we especially recommend the use of pro-grade lenses with HR mode(s)'), on this subject, as they can be often more self serving than accurate, or practical.

In the past, Olympus's own (IMO) flawed marketing has, I think, caused much more controversy, than a different marketing approach might have.

Just a thought.

I have found (subjectively) quite clear and distinct improvements from both T-HR and HH-HR modes in combination with 'the worst' lenses. The level of quality improvement (compared to 'the best' lenses) may actually be quite similar in practical terms, if a sense of relativity is kept in mind when forming one's judgement.

For example, some recent samples of HH-HR output, with the lowly 9mm f/8 bodycap lens:



-- hide signature --

"The camera introduces us to to unconscious optics as does psychoanalysis to unconscious impulses"
"The art of the critic in a nutshell: to coin slogans without betraying ideas. The slogans of an inadequate criticism peddle ideas to fashion."
- Walter Benjamin
"Drawing is a constant correcting of errors, maybe a great deal of creation is exactly that."
- John Berger
" photograph is to frame, and to frame is to exclude."
-- Susan Sontag

 Jan Chelminski's gear list:Jan Chelminski's gear list
Olympus E-M1 II Olympus OM-D E-M1X Olympus 25mm F1.2 Olympus Zuiko Digital ED 50mm 1:2.0 Macro Olympus Zuiko Digital ED 7-14mm 1:4.0 +4 more
Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow