The OMD photo thread

12-40

Scottboarding wrote in post #18482049
Absolutely gorgeous! How's the quality with the pixel shift? Does it really improve quality, or just create more pixels (if that makes sense)? I'm starting to really consider getting an E-M5 II for video and the pixel shift but if the pixel shift isn't all that great, I'll skip it.

Thanks.
My take is that YES the pixel shift is very useful - the comparison of the 20 MP "normal" to the shifted images is quite impressive. There's not much more in the 80 MP version than the 50 MP in camera jpeg, but the 80 MP allows raw processing and then you can scale it down to whatever suits the need, for printing for example. The expected caveat is moving objects - including things like water ripples - these can give rise to "herringbone" type artifacts - more pronounced with slower shutterspeeds, of course.
You be the judge. I thought so from the get-go but you prompted me to do a little more evaluation. I did a quick (and thorough enough for me) test this morning. The pixel shift option provides three files (if you are set to save L-fine jpeg + Raw) - a 20MP raw file, as normal; a ~50MP in camera jpeg and a 80 MP raw for when you feel the need to process the raw to get the best image. Sometimes in camera jpegs are good enough, sometimes the raw is needed to maximize the IQ for printing etc. Everyone likes different targets so I have a composite. These were all at f/4 MF with 1:1 focus peaking (the centre of the Siemens star makes that very accurate) ~1/40 tripod after 2 sec delay. Default processing (jpeg in camera and raw both set to no sharpening), assembled in PS and saved at 80 quality under save for web.
The target has some simple text lines, a Siemens star from Bart Van der Wolf and my print of Seattle Japanese garden (from 4 5DsR frames into a pano)
The first image below shows the whole frame and the yellow box is the crop area. It is based on scaling things back to 20MP (in PS) to see how much the 20 mp output can be improved by pixel shift/oversample and downsize. I think the colour purists may appreciate the sampling that means each final pixel in the image has info from R B and 2 G pixels so the Bayer issue goes away. I didn't address that, just overall appearance.

The second image shows a higher resolution shot of the target (which is pretty similar to the file for the print) and the crops show what everything looks like when upscaled to 80 MP for comparison. Just be sure to view at 100% otherwise the Moiré from the Siemens star will not be displayed properly...

If you are interested, I used Bart van der Wolf's calculation from the Siemens star to find the 20MP image to resolve 115 cycles/mm, the 50 was 188 and the 80 was 205, and the limiting frequency was 0.38 cycles/pixel, 0.39 and 0.33 respectively (@ Nyquist it would be theoretically 0.5). This is for the combination of lens and sensor but there's no AA filter. So the 80 only has a bit more than the 50 in terms of recorded cycles per mm sensor: it's hard to see much difference between 180 and 205 but there is clearly more detailed than 115!

Not an OMD, but PEN-F & 17mm 1.8

AJSJones wrote in post #18482682
Thanks.
My take is that YES the pixel shift is very useful - the comparison of the 20 MP "normal" to the shifted images is quite impressive. There's not much more in the 80 MP version than the 50 MP in camera jpeg, but the 80 MP allows raw processing and then you can scale it down to whatever suits the need, for printing for example. The expected caveat is moving objects - including things like water ripples - these can give rise to "herringbone" type artifacts - more pronounced with slower shutterspeeds, of course.
You be the judge. I thought so from the get-go but you prompted me to do a little more evaluation. I did a quick (and thorough enough for me) test this morning. The pixel shift option provides three files (if you are set to save L-fine jpeg + Raw) - a 20MP raw file, as normal; a ~50MP in camera jpeg and a 80 MP raw for when you feel the need to process the raw to get the best image. Sometimes in camera jpegs are good enough, sometimes the raw is needed to maximize the IQ for printing etc. Everyone likes different targets so I have a composite. These were all at f/4 MF with 1:1 focus peaking (the centre of the Siemens star makes that very accurate) ~1/40 tripod after 2 sec delay. Default processing (jpeg in camera and raw both set to no sharpening), assembled in PS and saved at 80 quality under save for web.
The target has some simple text lines, a Siemens star from Bart Van der Wolf and my print of Seattle Japanese garden (from 4 5DsR frames into a pano)
The first image below shows the whole frame and the yellow box is the crop area. It is based on scaling things back to 20MP (in PS) to see how much the 20 mp output can be improved by pixel shift/oversample and downsize. I think the colour purists may appreciate the sampling that means each final pixel in the image has info from R B and 2 G pixels so the Bayer issue goes away. I didn't address that, just overall appearance.

The second image shows a higher resolution shot of the target (which is pretty similar to the file for the print) and the crops show what everything looks like when upscaled to 80 MP for comparison. Just be sure to view at 100% otherwise the Moiré from the Siemens star will not be displayed properly...

If you are interested, I used Bart van der Wolf's calculation from the Siemens star to find the 20MP image to resolve 115 cycles/mm, the 50 was 188 and the 80 was 205, and the limiting frequency was 0.38 cycles/pixel, 0.39 and 0.33 respectively (@ Nyquist it would be theoretically 0.5). This is for the combination of lens and sensor but there's no AA filter. So the 80 only has a bit more than the 50 in terms of recorded cycles per mm sensor: it's hard to see much difference between 180 and 205 but there is clearly more detailed than 115!

***************
***************
***************

THANK YOU! I simple "yeah it's great!" would've been perfectly sufficient, but I really appreciate the effort! I never really though about the downsizing benefits; I just looked at it as a way to be able to make bigger prints. The color surprised me too. There seems to be more gradients of colors with the pixel shift which makes sense as you point out that it negates the Bayer part of the sensor.

Thanks for the detailed response and photos, I can't wait to see more! I have some shopping to do!

Scottboarding wrote in post #18482891
THANK YOU! I simple "yeah it's great!" would've been perfectly sufficient, but I really appreciate the effort! I never really though about the downsizing benefits; I just looked at it as a way to be able to make bigger prints. The color surprised me too. There seems to be more gradients of colors with the pixel shift which makes sense as you point out that it negates the Bayer part of the sensor.

Thanks for the detailed response and photos, I can't wait to see more! I have some shopping to do!

I had been thinking of doing a more detailed test than my first simple comparison (Yeah - it’s great ) when I got the camera, so you just gave me the nudge I needed - so, thank you Automated focus bracketing, time-lapse, in-camera HDR twos styles, lots of AE bracketing choices. Pro-capture, 10-15 fps shooting, silent mode etc etc - the pixel shift is one of many neat features - I’m still in awe of the 24 x 30” prints I made from those images (that’s a 44x geometrical enlargement!!). My “big” camera is the Canon 5DsR these days (used to be an Ebony 4x5) but Canon is sooo sloooow in adding useful features to the processor side

Really pushing the limits of the E-M1:

Tennessee

Fort Macon

PEN-F & 14mm 2.5

From the Brisbane City Hall's Clock Tower. - Trying to decide which one I prefer.

IMAGE LINK: https://flic.kr/p/21vb​BjW

IMAGE LINK: https://flic.kr/p/21vd​x7y

meno 2 ev e come da lettura esposimetro omd10

f 2,8

donnas

rod nait

PEN-F & Panny 14mm 2.5

oops..

Image hosted by forum (889404) © Robert Germans THIS IS A LOW QUALITY PREVIEW. Please log in to see the good quality stuff.