2019 - P52 - 4th Quarter - David (Rufus) - Merging / using multiple images

[Rufus]

For the 4th quarter of this year’s P52 I am going to explore combining images into a single image. Potential uses might include:

(a) merging images with different shutter speeds (HDR)
(b) merging images with different focus planes (focus stacking to increase depth of field)
(c) merging images with the same settings with the intention of “painting out” moving objects

I have not really attempted these techniques so far, so please bear with me, especially this week when, in the short time available to me, I have learned a fundamental lesson! But at least I have found my way around the relevant parts of the camera menu.

All comments will be welcome but I am more interested in learning about and practicing the techniques this quarter than looking for great compositions.

I have started with an HDR image and immediately learned that HDR is detrimental (and pointless) if a normal exposure can capture a reasonably exposed image! Here is the comparison of the HDR and “normal" images. A tripod was used throughout. The HDR merging was performed in Lightroom as I read it is better than Photoshop's equivalent.

#88 Gourds On The Vine (7 images merged):


#88 Gourds On The Vine (single image):


The "normal" image has better detail in the far walls and has not benefited the gourd plants where I thought the lights and darks might benefit. Contrast adjustments might have been more appropriate here, if needed.

[Manfred M]

(b) merging images with different apertures (to increase depth of field)

Are you sure about this technique? The standard approach to increase DoF is focus stacking. One stacks a number of images where the focus plane is shifted, but all are taken at the same aperture.

[Rufus]

Are you sure about this technique? The standard approach to increase DoF is focus stacking. One stacks a number of images where the focus plane is shifted, but all are taken at the same aperture.

Oops...that is embarrassing. I was indeed thinking of focus stacking. I will edit my original post.

[Geoff F]

I frequently use varying adjustments for 'exposure stretching' and increasing the focus depth, David, but I have never found an auto HDR programme where I have been completely happy with the results. So I always do my own edits with masking; which is sometimes known as 'hand made HDR'.

Where suitable, I take bracketed exposure shots then, after basic editing to the Raw images, I auto align the images as layers then use masks to edit in or out of any problem areas to produce a final image to my liking.

One of the problems with HDR is when there is subject movement, but with the hand merged method it is sometimes possible to work around this issue. For example, a person walking along a street where you want to improve the sky and reduce foreground shadows. As long as the person and the area behind them remains together, subject movement doesn't matter; as long as you have sufficient shutter speed etc.

My other frequently used technique is handy where the subject movement cannot be contained within a restricted area. I shoot one image then convert the Raw file in two or three ways which can be merged into the final scene. For instance, convert one image to suit the midtones, one for the highlights and one for shadows. Those images will always be perfectly aligned because they came from a single source. So they simply require stacking as layers with masks which can be edited. Not true HDR, more a case of 'exposure stretching', but it is usually a better way than attempting to push and pull just one image in different ways.

This will work well for hand held shots since you are starting your editing with a single image; although I often do HDR merges from hand held shots. Just make sure you have plenty of wasteable area around the edges for a final trim to remove misalignment areas.

If using Adobe Photoshop, this is an instance where Smart Objects can prove useful. Simply open the original image, after basic editing, as a Smart Object and copy extra images which can easily be returned to ACR for extra editing in any way required. Smart Objects do require a little bit of initial thought about their operation but once learned this method becomes 'second nature' and is very easy to use.

[Rufus]

These are interesting approaches approaches, Geoff, which I may well try once I have become more familar with the more automated methods. Thank you for your detailed account as I recall you often mention in your threads that your images are a blend of mutliple handheld images images with diferent exposures I have stored the mothods for future reference.

Whichever method, I must remember to "auto align" which I completely forgot about on this first attempt, although using a tripod on this occasion may have mitigated any need for it.

Thank you again for commenting.

[DanK]

I shoot one image then convert the Raw file in two or three ways which can be merged into the final scene. For instance, convert one image to suit the midtones, one for the highlights and one for shadows. Those images will always be perfectly aligned because they came from a single source. So they simply require stacking as layers with masks which can be edited. Not true HDR, more a case of 'exposure stretching', but it is usually a better way than attempting to push and pull just one image in different ways.

I have to disagree with the way Geoff characterized this. It isn't better than pushing and pulling one image; it's another way to push and pull one image. It's another way to do local adjustments to a single image. I know some people find this a better method than doing those local adjustments in other ways--e.g., by using different masks on dodging and burning layers--but it's still just adjusting one image.

The only reason I fuss with HDR or an alternative is if there is a problem with the single image that a second image will solve. The most common reason is a scene that exceeds the dynamic range of the sensor, but there are other reasons as well. E.g., suppose that an image has a lot of poorly exposed shadow,just barely above the exposure floor for that sensor. That part of the image will have a low signal:noise ratio, so you can get a better image by merging that region from a second image with a brighter exposure than by lifting the shadows in postprocessing.

There are several options for doing this. One is by hand: take two or more differently exposed images, stack them as layers, and use masks to select well-exposed regions from each. A second is exposure fusion, which is close to an automated version of the previous. The software simply selects regions from each image that are well exposed and merges them into a composite. The third is HDR.

With one exception, I never use HDR software for this because I dislike the distorted colors and other artificial-seeming patterns it can create. My go-to method for a long time was a Lightroom plugin for exposure fusion called LR Enfuse. The sole exception is that recently I started to use use the merge to HDR function in Lightroom quite often because, much to my surprise, it has yielded results very similar to exposure fusion, without obvious distortions. Maybe I have just been lucky so far.

In many cases--certainly with exposure fusion and LR's HDR merge--the resulting image often needs a lot of work on tonality. The transitions from light to dark are often unnatural and drab. LR has an automatic tonality adjustment that sometimes gives a serviceable starting point.

[xpatUSA]

For the 4th quarter of this year’s P52 I am going to explore combining images into a single image. Potential uses might include:

(a) merging images with different shutter speeds (HDR)
(b) merging images with different focus planes (focus stacking to increase depth of field)
(c) merging images with the same settings with the intention of “painting out” moving objects

I have not really attempted these techniques so far, so please bear with me, especially this week when, in the short time available to me, I have learned a fundamental lesson! But at least I have found my way around the relevant parts of the camera menu.

All comments will be welcome but I am more interested in learning about and practicing the techniques this quarter than looking for great compositions.

I have started with an HDR image and immediately learned that HDR is detrimental (and pointless) if a normal exposure can capture a reasonably exposed image! Here is the comparison of the HDR and “normal" images. A tripod was used throughout. The HDR merging was performed in Lightroom as I read it is better than Photoshop's equivalent.

#88 Gourds On The Vine (7 images merged):


#88 Gourds On The Vine (single image):


The "normal" image has better detail in the far walls and has not benefited the gourd plants where I thought the lights and darks might benefit. Contrast adjustments might have been more appropriate here, if needed.

Were the 7 images prepared in any way before merging (e.g. alignment, weeding out ones having movement, etc.)?

You could say that, if contrast adjustments actually work, HDR would not have been needed, IMHO. In other words, HDR stacking and merging is for those scenes where contrast adjustment of a single image would not work, i.e. scene DR exceeds camera DR.

I use the technique (align_image_stack.exe, enfuse.exe, Hugin, TuFuse Pro, etc.) occasionally but lose patience quite quickly ...

I hope your quest goes well, don't let me discourage you!

[Rufus]

You could say that, if contrast adjustments actually work, HDR would not have been needed, IMHO. In other words, HDR stacking and merging is for those scenes where contrast adjustment of a single image would not work, i.e. scene DR exceeds camera DR.

We are on the same page, Ted, but I think you have put it far better than I did.

I used a tripod. It was not a windy day and the only movement was of people behind the "vine" but I mitigated this by waiting between shots so that they were mainly hidden, if not fully out of the frame. Consequently I did not weed out any shots. And I completely forgot about alignment, but using a tripod should have reduced the effect of that omission.

I have now found out how to change the number of bracketed shots, so I likely will not normally use as many for exposure bracketing.

I did not let Lightroom apply my normal automatic adjustments on import as I thought I should leave the images as taken and then adjust the merged result. But I see Geoff does some adjustments prior to applying his "manual HDR" method.

It is extremely early days so I am not discouraged just yet

[xpatUSA]

We are on the same page, Ted, but I think you have put it far better than I did.

It is extremely early days so I am not discouraged just yet

Thanks. OT, but I love that Victorian (?) behind the wall, especially the patterns and the Engineering bricks. I see the gargoyle on the ridge - is it a church or community hall?

[Manfred M]

I have tried a number of different methods of handling high dynamic range scenes and the one that was most effective was to buy a camera that has a very high native dynamic range and then locally adjust those areas to show detail. The only time I tend to use multiple exposures is when my camera histogram shows data loss in both the highlights and shadow areas. This approach is conservative because I don't know if I truly have blocked shadows and clipped highlights in the raw data because the camera's histogram is calculated on the JPEG.

If I do see clipping in those scenes, I will generally take at least three shots; -2, 0 +2. If I still see clipping on the histogram, I will go further and shoot -4, -2, 0, +2, +4. I rarely have to shoot more to capture a full dynamic range but it has happened on occasion. Sometimes I shot one stop increments (when I was shooting the Nikon D800 the camera allowed a maximum of 1 stop increments, the D810 allowed 2 stop).

I have tried three different methods; HDR, exposure fusion (using Enfuse / Enfuse GUI) or hand blending the layers.

Here is a shot I used all three methods on; I did a -2, -1, 0 , +1, +2 hand-held shot here. The shot was taken in a Buddhist temple in the Himalayas. There was no artificial light, so the room was quite dark except for some windows. The side windows were so bright that they blew out in a single exposure and the shadows on the opposite side were so dark that there was no shadow detail.


The five images I worked with:




This first image was produced using Adobe's HDR algorithms using a classic tone mapping approach:






This second image was produced using Enfuse (http://software.bergmark.com/enfuseGUI/Main.html. Enfuse selects specific areas of the images and does not tone map, but rather uses areas out of each of the images without changing the colours or brightness and builds a new image from these parts of the images.






In this image, I stacked the images together in Photoshop and used hand-painting techniques on the layers to reveal details that I wanted, using the middle "properly" exposed image as the base to work from. This is somewhat along the lines of how Enfuse works except that I chose how to blend the various images together.

[Geoff F]

The differences are quite clear and, for me, the last one is definitely the best result.

[DanK]

when you say Adobe's HDR algorithm, which one? If I'm not mistaken, the one in Lightroom is not the same as the one in Photoshop. The latter, when I tried it some years ago, created noticeable color distortions. I haven't yet noticed them with the LR algorithm.

I'll put on my mental to-do list trying some similar comparisons with LR, if you used Photoshop.

[Manfred M]

when you say Adobe's HDR algorithm, which one? If I'm not mistaken, the one in Lightroom is not the same as the one in Photoshop. The latter, when I tried it some years ago, created noticeable color distortions. I haven't yet noticed them with the LR algorithm.

I'll put on my mental to-do list trying some similar comparisons with LR, if you used Photoshop.

I used the Photoshop one (rather than the ACR one that is identical to Lightroom). The newer ACR / LR one is often viewed as being superior to the Photoshop one, but I find I generally prefer what PS does.

[Rufus]

Thanks. OT, but I love that Victorian (?) behind the wall, especially the patterns and the Engineering bricks. I see the gargoyle on the ridge - is it a church or community hall?

I like the bricks as well, and the way the courses are laid.

There is a road behind the wall and on the other side is a a domestic rather than a community property. The sculpture on the end of the roof's ridge is known as a roof finial. It is purely decorative and the sculpture is melded with the ridge tile during manufacture into a single item.

[xpatUSA]

I like the bricks as well, and the way the courses are laid.

There is a road behind the wall and on the other side is a a domestic rather than a community property. The sculpture on the end of the roof's ridge is known as a roof finial. It is purely decorative and the sculpture is melded with the ridge tile during manufacture into a single item.

Thanks, hadn't heard of them until now. Found this:

https://www.house-design-coffee.com/...s.html#gallery[pageGallery]/1/

Pretty rare on my property with it's all-tin roofs ... best I've got is a piece of tin angle ridge-cover sticking out a foot or so.

[DanK]

Here's the test pair from years ago that led me to give up on whatever HDR was in photoshop at the time.

Here's the image composited with exposure fusion, using the Lightroom enfuse plugin. The colors are accurate.




Here's the same thing composited with Photoshop's HDR:




This a good test image because it includes one color everyone can use as a reference without having been there: he sky. Note that the HDR sky is an unnatural turquoise. The colors of the rocks and lilies are also off.

[Rufus]

I agree with Geoff that the third is superior to the others. Unlike you, the algorithms don't know what is the subject and how much or how little to lighten the background to bring the viewer's attention to the subject or to model the light on the subject. As usual, you have made an excellent job of it, Manfred.

[Rufus]

From some initial internet searches it seems that to manually blend multiple images that have differing exposures requires the use of blending modes, chanels and, ideally, smart objects, all things I have heard of but am not yet familiar with.

In the short term it will be easier to use the Lightroom HDR option, until disatisfaction forces my hand.

[xpatUSA]

From some initial internet searches it seems that to manually blend multiple images that have differing exposures requires the use of blending modes, channels and, ideally, smart objects, all things I have heard of but am not yet familiar with.

David, I see that the above is mostly written in Adobe-speak. Have you thought about other, possibly easier-to-use, non-Adobe apps?

In the short term, it will be easier to use the Lightroom HDR option until dissatisfaction forces my hand.

The best of luck to you. We standby to help the best we can ...

[Manfred M]

David, I see that the above is mostly written in Adobe-speak. Have you thought about other, possibly easier-to-use, non-Adobe apps?

Blending modes and channels are not "Adobe speak", but SmartObjects are. Channels are colour channels, something that is we see in many other editing programs. Blending Modes (called Mode in GIMP) are also found in many pixel based editors (Affinity Photo, Corel Painter Pro, etc.).

SmartObjects / SmartFilters are a wrapper that are (unfortunately) unique to Adobe.