Across Riddlesdown, at 14mm

Shoot you? I assume you are volunteering a portrait image as a second test?

Thoughtful of you. Important to replicate these kinds of studies with different types of images.

Feel free to load up the RAW image at your convenience, Matt. No need to wait for someone to come and take a snap of you.

No wonder we were getting different results then
Unfortunately I didn't get that to work from - it would have made a better starting point I think. Your version from the 'darker' raw looks very similar to some of the renderings I was getting. What I couldn't get was the extra cloud detaiul you have from the 'darkest' one - that mystery now being solved as we now know it was there in your raw but not mine

On my monitor, the "darker" and "darkest" versions that I posted above are visually almost identical. Any slight tonal differences are down to me and not the source DNG or software used.

I did follow on this morning and generate a merged version using the LR/Enfuse plugin in Lightroom - but the results are so similar to those from single DNG sources that it's hardly worth posting. It is, though, impressive (to me) that LR/Enfuse could blend the source images with no noticeable degradation.

My conclusions remain the same. If you get a useable RAW with (relatively) unclipped highlights, then taking several images and using HDR software to merge them is unlikely to improve the result compared with optimally processing that single RAW. I can imagine circumstances where that might not be true - but for the images in this thread, I think it is.

As far as the image itself is concerned, I think it suffers from not having any obvious subject. The path in the foreground leads the eye to some rather uninteresting trees in the mid-ground. HDR does not produce a better result - but pointing the camera elsewhere might.

Andrew

There's some impressive analysis going on here! Those screen shots are very amazing showing all that data, thanks guys. This does focus the issue only our exposure options and choices, and its very interesting to be gaining more insight into the subtlies behind the ETTR approach and what software can drag out from an image.

Just to clarify, there was some confusion over the DNGH files earlier - Rob originally asked me for an example representing an 'ETTR' type exposure, so I gave him the middle bracket plus 1 higher - these I named 'Darker & Brighter'. I understand the brighter has proven faily unusable on its own.

Then Andrew asked specifically for the lowest exposure bracket (presumably becasue it has become clearer that the sky is most problematic to recover), so I gave him the the next bracket down - so this I called 'darkest'. This is now showing us some impressive shadow recovery capability, even from the lowly K7. But is the IQ anything the equal to the HDR alternative?

On mine I see a lot of very desirable extra detail in the clouds from your 'darkest' image With which I agree. I felt even with the 'darker' image source single shot results were sufficiently good that HDR processing yielded only minor benefit. And 'conventional' single shot processing from that 'darkest' image narrows the gap even further. I often think much the same thing when I review my own pictures

The scene was lit by quite diffused lighting through a lot of cloud cover, so yes, there was good detail visable in the deeper shadows under trees and in the distance. It was not a squinting sort of scene, the eye could scan around easily and adjust taking in a high quality of deatil - thats what I wanted to reproduce, what was natural able to be seen.

My own feeling is that this is better represented by the quite effortless HDR process, whereas other single file techniques rely on high 'oompph factor' software power, especially to drag detail out of deep shadows, the weakest parts of the sensor detail.

Nigel, I have emailed you a fullsize JPEG from the "darkest" DNG. I hope you find it interesting/useful.

I don't claim that it is the best possible result from that DNG using Lightroom - but it is close to the best that I can achieve in less than 5 minutes.

Andrew

I have examined the full resolution images in Lightroom at 100% and can see little or no difference between the cloud detail in the "darker" and "darkest" images.

Can you post crops from the areas of the two images, side by side so that I can see what you see?

Andrew

I'm just looking at the images you have posted here Andrew so I can't post 100% crops of those. I'm just saying that I can see cloud detail in your 'darkest' source images for areas where both your and my 'darker' source images display no detail as the original raw is clipped for these areas. If Nigel should send me the 'darkest' raw I could do some 100% comparisons - although I'm running out of enthusiasm to spend more time on this.

These are 100% crops from the full size images that were used to export the JPEGs posted earlier. You said that you could see "very desirable extra detail in the clouds" - so where is it?




The version on the right is from the "darkest" DNG.

Andrew

I see absolutely no difference at all here on my monitor at work, but it's hard enough to see anything at all on that so no surprises there I'll have to have a look when I'm back on a decent calibrated monitor.

These are all valiant & superb efforts. What I do find curious, and a bit ironic, is the extraordinary commitment to the single file option .... originally this debate kicked off with comments about the non-necessity for HDR, and how much more trouble it was - the word 'hassle' was used to describe it. This seemed to be mostly accepted. But here we are still trying to prove that we can find one ideal single exposure and process it to perfection .... and it has taken many pages of discussion and describing of software techniques .... and still its all in the name of finding something that equates to or nearly equates to the original HDR!!! I may well still argue its not there yet.

Where is the real hassle?? As I said earlier, truth is, the creation of the HDR (both the bracket shooting and the Photomatix work) was painless - no tripod required and the processing was easily sorted with the 'natural' Tone Compressor preset and a few small tweaks here & there. Even NR, aberration correction, de-ghosting and image alignment was taken care of automatically while I sat back for about 45secs.

Even if I had one single perfect exposure it would take far more effort to optimize it, involving inevitably a big compromise somewhere to strike the right balance between sky & ground. From what Andrew & Rob have been demonstrating, it would seem that in order to get an acceptable single exposure of this scene we would need to shoot at about -2.5 stops compensation in order to get a good sky, producing a near black ground in the process, and then crank up all the shadow recovery and lifting processing oomph we can muster out of Lightroom or PS, just to return the ground to something we can see properly again.

How can this be any good for IQ? The camera sensor records most of its quality data in the upper half of the spectrum - thats the rationale behind 'ETTR'. By underexposing most of the scene are we not throwing away so much of the best quality data the camera could record?

The HDR technique gives us the best of it all - each bracketed exposure is optimal for a certain band of DR, right from the darkest (which gives us the best sky highlights here) through to the brightest (which gives us the shadow areas smack in the middle of the sensors sweet spot.) The software does most of the rest, along with the users careful and subtle control.

I remind you all - no one knew this image was HDR until I said so. Then all of a sudden it was 'no need for HDR for this scene, one shot would do it, etc etc'. Does it seem so clear cut now?

Yep

Nope.

The highlight detail that was correctly exposed (in the "under exposed" image) would be used by the HDR software anyway. The area where the HDR software might be expected to gain (compared with using only your "darkest" image) is in the shadows, where a few percent of the darkest pixels have an RGB value of zero in your "darkest" DNG - but that isn't visually obvious.

Andrew

Very clear cut. If you prefer your HDR images - then use HDR. I don't, so I won't - unless the scene demands it's use, but that is relatively very rare.

Simples ......

Andrew

Nigel, the images were generally uninspiring, maybe if they had been, they would have had additional scrutiny. I know for my part, I was just looking for someething positive to say