Test EM1.3 (and EM1x) HHHR - DR improvement

[pdk42]

I do a lot of landscape and find myself often running out of shadow lifting capability (i.e. before noise becomes an issue). I often resort to shooting a 5-shot, 5-stop bracket and stacking later in LR (I always shoot raw). This works pretty well, but it consumes a chunk of storage and takes time to stack on the PC.

I knew that HHHR could perhaps provide an alternative since it works by additive stacking of multiple images and this definitely reduces noise/improves SNR. In fact, you can do HHHR yourself without a EM1.3 simply by firing a burst of images at the same exposure and then using Photoshop or similar to do a "mean" stacking. But this takes even longer than HDR stacking so it's not something that's sensible for regular use.

So, I thought that getting a mk iii would be perhaps be worth it for HHHR. To prove it to myself, I did some playing with HHHR test images I found on the web and this persuaded me that the technique would offer me something worthwhile. I'm not in the least interested in the extra resolution - just the DR. To add weight to my thoughts, Bill Claff's excellent photons to photos site provides this nice DR comparison chart:

Subscribe to see EXIF info for this image (if available)

At nearing 12-stops, it's pretty much in line with FF cameras. So, I bought the mk iii.

To test out the HHHR myself, I took the shot below. Exposure is deliberately set so as to avoid burning out the sky. According to the meter, it was -3EV. It's a bit of a torture test given just high the DR was (cloud strongly backlit by the sun and the foreground in deep shadow):

Subscribe to see EXIF info for this image (if available)

EM370001-3_5184_ooc by Paul Kaye, on Flickr


I then pushed shadows 100%, then added +2.2 EV to the bottom part of the image. Excuse the very crude masking - it's just to show the principle. First shot is the HHHR version. Second shot is the std res version (.ORI file):

1. HHHR res file, pushed 100% shadows and +2.2EV on the bottom of the image

Subscribe to see EXIF info for this image (if available)

EM370001-3_5184 by Paul Kaye, on Flickr

1. Standard res file, pushed 100% shadows and +2.2EV on the bottom of the image

Subscribe to see EXIF info for this image (if available)

EM370001L-2_5184 by Paul Kaye, on Flickr

So, already a big difference. in terms of colour and clarity. I've uploaded these as full-size shots so if you want to look at them full-size, just download them - but note that I've down-resed the HHHR version to the same size as the standard res one - I'm after HHHR not for the extra resolution, but for the improvement to S/N ratio.



I then shot a 5-shot HDR sequence (-2, -1, 0, +1, +2) with the meter reading -1.5EV - in other words to allow the -2 EV shot to be comfortably not over-exposed. I shot 5 raw files and stacked in LR. Here's the result:

Subscribe to see EXIF info for this image (if available)

EM370002-2-HDR_5184 by Paul Kaye, on Flickr

This is certainly better than the HHHR version, but OTOH, it's slower to capture so foliage can get ugly in a breeze, it chews up a lot more storage, and needs quite a bit of post-processing to build. Also, the scene here is pretty extreme, so I think in many cases, the HHHR shot will be a sensible compromise.


The crops below show the noise off the various shots:

1) HHHR shot.

Subscribe to see EXIF info for this image (if available)

2) Sts Res shot:

Subscribe to see EXIF info for this image (if available)

3) HDR shot:

Subscribe to see EXIF info for this image (if available)

So - HHHR does definitely deliver better shadow noise and therefore better DR. For something that's been pushed over 2 stops in PP, plus 100% shadow fill, it's actually pretty impressive. I think it'll make a real difference to tricky landscape shots - but it's not as effective as a HDR bracket.

So, it's a win then? Well, it's certainly a useful tool, but you do need to balance these negatives:

- I tried it today on a windy day and foliage movement renders the technique pretty useless - the blurring is way beyond acceptable. OTOH, it's less ugly than a HDR stack.
- The time it takes to do the stack is about 15-20s. That doesn't sound much, but it definitely breaks the flow.
- Sharpening of the raw is tricky. I think I'm there now, and it's just a preset in LR, but even so...

Overall, I'm glad I have it, but I'm not convinced it's a feature that's worth the upgrade cost alone.

 

[stevedo]

I tried to click through to the photos on Flickr but they seem to be private.

Subscribe to see EXIF info for this image (if available)

 

[pdk42]

I tried to click through to the photos on Flickr but they seem to be private.

Ah, yes. I guess it will be. Actually, you don't need to click through to see the images in full res since I uploaded them full-sized anyhow. If you want to view them full-sized locally, then right click and save them to your PC.

 

[retiredfromlife]

Pretty impressive to me. I think I would use the HHHR just for convenience.
Wonder how or if it would work for macro or just close up photography

 

[comment23]

Pretty impressive to me. I think I would use the HHHR just for convenience.
Wonder how or if it would work for macro or just close up photography

You might want to take a look here: https://smallsensorphotography.com/hi-res-cactus-images/

here: https://smallsensorphotography.com/floral-showhouse-macro-macro/

and here: https://smallsensorphotography.com/hhhr-macro/

 

[0000]

Actually, you don't need to click through to see the images in full res since I uploaded them full-sized anyhow. If you want to view them full-sized locally, then right click and save them to your PC.

The forum software resizes everything to 1600px on the long side, regardless of the size uploaded, so that won't work.

 

[MichailK]

a minus point for me into getting the e-m5iii - damn, Oly! They should have added a co-processor/accelerometer option in the optional grip option just for that! It would still be much smaller and not that ”pro” to steal em1iii sales and no, I am not getting in the em1iii size whatever the price

 

[pdk42]

a minus point for me into getting the e-m5iii - damn, Oly! They should have added a co-processor/accelerometer option in the optional grip option just for that! It would still be much smaller and not that ”pro” to steal em1iii sales and no, I am not getting in the em1iii size whatever the price

Well, you pays your money...

I do like the E-M5 line size though. The E-M1 line is still acceptable to me though.

 

[bassman]

I ran a different dynamic range test in (my attempt at) controlled lab conditions, and only at base ISO (200). Rather than shoot a scene with a large dynamic range and try and see what I could recover, I shot a middle grey card (actually a cardboard box) on my E-M1.3 in three modes:

1. On a tripod in normal mode
2. On a tripod in HDR tripod mode
3. Handheld in HDR handheld mode.

Subscribe to see EXIF info for this image (if available)

All shots were f/4, ISO 200 using the 12-100/4 lens at 66mm. From an exposure that yielded middle grey (where the camera said I was exposed "correctly"), I varied the shutter speed faster and slower in full stop increments. I deliberately defocussed the lens to smooth out the imperfections in the surface I was shooting. It was lit by overcast light through a window. I viewed the images in LR Classic after reducing Saturation to -100 and used the RGB values along with visual inspection to determine when the image went to all black or all white. Since the lens has some vignetting and the window light also tailed off away from the window, I scanned the central region of each image and took my consensus reading of the RGB values for that area.

In Normal mode, the "correct" exposure was 1/30s. This yielded an RGB value of about 50%. The shortest exposure showing any visual difference from pure black was 1/500s at 1.5%, with 1/1000 at 0.1%. The longest exposure showing any visual difference from pure white was 1/2s at 99%, with 1s being white at 100%. So 11 stops from pure black at 1/1000s to pure white at 1s.

In HDR tripod mode the "correct" exposure was 1/60s at RGB 47%, The shortest and highest exposures which weren't blown were 1/1000s and 1/4s. So still 11 stops from black at 1/2000s to white at 1/2s.

In HDR handheld more the results were similar to the tripod mode. The RGB values were slightly shifted at each EV, but not enough to make a difference. For instance, the "correct" exposure was still 1/60s with RGB at 45%. So again, 11 total stops from black at 1/2000s to white at 1/2s.

Other than the shift of one stop in the range, it appears that the size of the range is unchanged by HDR - it's eleven stops from black to white, with middle grey falling squarely in the middle of the range. You can get by with one stop less exposure and get the same brightness in your images.

 

[bassman]

Just for yuks, I constructed my own Zone Chart using the images I shot for this test.These are the 9 non-blown strips.

Subscribe to see EXIF info for this image (if available)

Being a noob in PS, it took me a bit to figure out how to make it work. But it was fun, and now I know how.