Req: Better HDR exposure merge

[photoken]

PL's Create-HDR Image leaves a lot to be desired. Truth be told, it's unusable.

For comparison, I used the 5 sample CR2 RAW images available from this Web site:
http://captainkimo.com/hdr-software-review-comparison/

Here's PL's result:

HDR test PL.jpg

Here's the result from Paintshop Pro (without any additional adjustments):

HDR test PSP.jpg

Fortunately for PL, it can open images created by HDRMerge, so we can have this as a starting point:

HDR test HDR Merge in PL.jpg

[greenmorpher]

I have never had any joy with HDR in PL, which is a pity since it had HDR (which the brothers chose not to call HDR) well before PShop did.

Now, my Panny G6 does a much better job of HDR than PL does.

Cheers, geoff

[Gerhard Huber]

The HDR function of PhotoLine works with flat images only. Since you used Raw images and PhotoLine opens Raw images with layers, it doesn't work correct. Try to use the JPEGs and you will get much better results.
If the result is too bright, move the gamma slider to 0.

Gerhard

[photoken]

The HDR function of PhotoLine works with flat images only. Since you used Raw images and PhotoLine opens Raw images with layers, it doesn't work correct. Try to use the JPEGs and you will get much better results.
If the result is too bright, move the gamma slider to 0.

Aha! Thanks for the explanation -- using flat images is the key, all right. I didn't like the result using JPG, but using TIF worked quite well. (In the HDR editing window, I found that I needed to reduce the Contrast to zero to eliminate halos.)

For each CR2 image, I used the default RAW settings and only added a Chromatic Aberration correction before saving the image as a TIF.

Here's the result without any subsequent adjustments:

HDR test TIF in PL default.jpg

That's a good starting point. Adding three adjustment layers and playing with their settings yielded this:

HDR test TIF in PL adjusted.jpg

That result is very nice. The original image has more light in the shadows and more colour and details in the shadows, too, than these heavily reduced size JPG (for forum posting) samples show. I suspect that if I spent more time adjusting all the settings for the original RAW images before saving them as TIFs, the results would be even better.

I'll add some requests in subsequent posts in this thread.

[photoken]

This request is for the sake of convenience and ease of use -- it would be very good for PL to be able to use RAW images when creating HDR images.

As it is now, it's a very tedious and time-consuming process to first open each RAW image and save it as a 16-bit TIF image before loading those TIFs into the HDR processor. When PL opens a RAW image in the Editor, it applies an automatic correction to it anyway, so why can't the HDR feature load the RAW images with that automatic correction applied and then flattened?

That ability would put PL on a par for convenience with applications like Paintshop Pro, which can use RAW images directly in their HDR exposure merge utility. One would still be able use the current method to first enhance the RAW conversion before saving as TIF, if one wanted the ultimate control over the process.

[photoken]

When the HDR "Contrast" (gamma) slider is set at the default "1.00", that setting is unusable because of the halos around dark objects:

HDR halo.jpg

I have to reduce the setting to "0.30" or less to eliminate the halos, which significantly restricts the available gamma range. The resulting HDR images will be quite dark, making it necessary to use some extreme corrections in the post-processing adjustment layers to lighten the shadows and get good shadow details.

I don't know if those halos are an inherent part of generating an HDR image, but is it possible for the PL algorithm to suppress halos or eliminate them?

[Herbert123]

Hey Ken, you may want to try EnfuseGUI: http://software.bergmark.com/enfuseGUI/Main.html

I got this with the basic settings:

Untitled.jpg

And with some simple quick Curves in HIS mode:

Untitled2.jpg

Really nice result, I would say. EnfuseGUI does a great job.

[photoken]

Hey Ken, you may want to try EnfuseGUI:

Been there, done that. When I discovered HDRMerge (and once RawTherapee supported its images), I happily ripped Enfuse and all its associated software off my drive, shouting with joy! Now that I know how to use PL's HDR Creation tool and the appropriate adjustment layers, Enfuse and its ilk will certainly stay a long way away from my computer.

Really nice result, I would say. EnfuseGUI does a great job.

I don't like that result because of its overdone, phony look. For a long time I was skeptical about exposure merging to create HDR images because of that ugly look, until one of the RawTherapee developers explained to me that there's really two components to what's generally known as HDR imaging:

1. The creation of an image with good detail in both the lightest and darkest areas.
2. "Tone mapping", which has the effect of boosting the shadow values and increasing the general contrast.

It's that tone mapping which gives the overdone, phony, ugly "look" that has come to be associated with HDR images.

Of course, that "look" is a matter of personal opinion. If one wants that, it can be achieved in PL by creating an HDR image and applying a Curves adjustment in Lab mode and using an extreme curve on the Lightness channel:

HDR from TIF 01 tone mapped.jpg

I prefer a more "naturalistic" look for the HDR image:

HDR from TIF 01.jpg

The great thing about PL's HDR Creation tool is that both preferences can be accommodated with it.

[Herbert123]

I think it all depends on the look and the feel you are after - and it has to be decided on a per case basis. What works for one image, will not work for the next. And yes, I am aware of the technical aspects of shadow/light contrast versus tone mapping.

I agree with you that the fake HDR effect, when overdone, can kill the intent of any photo. One aspect I dislike with a vengeance are the halos often produced where dark and light touches - it kills the picture.

Although I have to say that in this particular case I find your versions a tad dull and flat to look at (even the overdone one). Your more realistic version looks extremely dark in the foreground, with nearly no visual detail visible. That is not how human eyes would perceive that scene.

Being a digital painter, designer, and 3d artist, I come from a different point of view in that an image has to provide visual interest - and sticking to "realism" (which is very much in the eye of the beholder) should not used as an argument to create boring uninteresting pictures with a lack of contrast.

Obviously this photo isn't that great to begin with. It is very average - and I think overdoing the colours, details, and contrast help to make it look a little bit more interesting.

The same is very much true for film: a tremendous amount of colour grading is done to the extent that, when one compares the result to the actual "real" shots, they look completely unrealistic. However, visual interest is increased, and the art choices made during colour grading assist to set the tone for that scene/movie.

In the end you ask yourself: do I want "realism" (which really does not exist for a photo, because our eyes and brains interpret reality, light, shadows, and so on, in a completely different manner than a camera lens and camera's tech), or do I want a captivating image that offers a lot of visual interest and details? Colour contrast or flat? Vibrant or dull. I prefer a visually exciting, yet unrealistic image, over a realistic image that looks uninteresting and is easily forgotten.

(Most) painters and artists never went for true realism, but for a lasting impression, and art is always a personal interpretation. Same with HDR and photos in my opinion. If anything, a photo is a lack-luster interpretation of reality - it only captures a small part.

Then again, with the right setup, lighting conditions, lense, camera, composition, and so on absolutely beautiful photos can and are shot every day by professional photographers without any need for HDR trickery. One could argue the setup and tech, and the way it is used in a shoot, provides for enough trickery by itself to "improve" the look of reality.

Anyway, I have used HDRMerge and other apps in the past - for some reason I missed out on Enfuse. I like the details it brings out in the shadows and highlights that I miss in Photoshop, Photoline, and HDRMerge. It also took me but a minute without any prior experience to get a visually interesting result, even with this mediocre photo, which was surprising to me (especially after having to deal with Photoshop's HDR merge disaster in the past). But perhaps I just have not played enough with the various settings.

[bkh]

For a long time I was skeptical about exposure merging to create HDR images because of that ugly look, until one of the RawTherapee developers explained to me that there's really two components to what's generally known as HDR imaging:

1. The creation of an image with good detail in both the lightest and darkest areas.
2. "Tone mapping", which has the effect of boosting the shadow values and increasing the general contrast.

It's that tone mapping which gives the overdone, phony, ugly "look" that has come to be associated with HDR images.

The image created in 1. is the true HDR image, of course. Problem is that there is no display device to cover that tonal range, so you have to resort to 2. (compress tonal range, i.e. reduce global contrast, trying to maintain colours and local contrast). Getting saturation right seems especially difficult.

enfuse uses a different approach, basically combining the best exposed parts of each input image. In the example, you can use a well-exposed image of the sky together with one of the foreground and merge them together using a mask, and you'll get essentially the same result. This maintains the colour and local contrast of the original photos, as long as large enough parts come from the same image – enfuse works best if you have two differently lit areas with no transitions. In principle, you can achieve the same using an HDR image and an appropriate tone mapping function, but this is not easy, so it's usually good to have both methods at one's disposal.

Cheers

Burkhard.

[greenmorpher]

Following the discussion here, I have now extracted a much better result from PL's HDR than previously. Of particular importance was Gerhard's oointer to shift the gamma from 1 to 0.

I wonder why it is set to 1 as default?

However, as good as the result might be, even with tweaking as suggested above, it still does not match the result I get from enfuse through Bracketeer with one click.

Cheers, geoff

[photoken]

The image created in 1. is the true HDR image, of course. Problem is that there is no display device to cover that tonal range, so you have to resort to 2. (compress tonal range, i.e. reduce global contrast, trying to maintain colours and local contrast). Getting saturation right seems especially difficult.

Increasing the saturation was the thing that took me the longest time to figure out in PL.

At first I simply tried a Hue/Saturation adjustment layer. The colours looked great when the image was displayed at Full View (21%), but zooming in to 200% revealed a blue "ghost" next to the darkest areas that was impossible to remove.

The I hit on the solution of adding a Curves adjustment layer, working in HIS mode, and raising the far left curve point of the Saturation channel. It's very sensitive to changes, so I used the up arrow key to nudge it up -- several clicks did the trick.

enfuse uses a different approach, basically combining the best exposed parts of each input image.

Seems to me that that is how every HDR creation tool, including PL, works.

[photoken]

One aspect I dislike with a vengeance are the halos often produced where dark and light touches - it kills the picture.

I couldn't agree more. That's why I asked for halo suppression/elimination to be incorporated in PL's HDR creation algorithm.

Although I have to say that in this particular case I find your versions a tad dull and flat to look at (even the overdone one). Your more realistic version looks extremely dark in the foreground, with nearly no visual detail visible. That is not how human eyes would perceive that scene.

Yep. Although much was lost in the translation from the original image to a 19% JPG (to allow posting in this forum), in general your comments about the dark foreground are correct. For this particular image I made two decisions, before modifying a single pixel:

1. The most interesting parts of the photograph are the clouds. Therefore, those are the areas that will be emphasized in the final result. Their reflections in the still waters of the pond are also nicely rendered.
2. The details of the grasses and shrubbery along the banks of the pond are uninteresting. Therefore, those areas will be de-emphasized. Those details will still be "there", but not obvious unless one stares at those parts of the image. In other words, I placed them in Zone II (if I remember my Zone System correctly ).

I think that if one were at the scene and looking directly at the setting sun and its surrounding clouds, the iris of the eye would be constricted to accommodate that brightness, so the details in the darker, shadowed areas on the banks of the pond would not be very apparent. At least, that's how I imagine one's eye would interpret the scene if one were there....

(The contrast can easily be boosted by using a Curves adjustment layer.)

In the end you ask yourself: do I want "realism" (which really does not exist for a photo, because our eyes and brains interpret reality, light, shadows, and so on, in a completely different manner than a camera lens and camera's tech),

Absolutely right. That's why you put the word "realism" in quotes and why I put the word "naturalistic" in quotes. In all the visual arts (not just photography), there's a continuum from the highly detailed, recognizable to the completely abstract. The great thing about Art is that each artist can present his interpretation of a scene wherever on that continuum he sees fit.

[bkh]

enfuse uses a different approach, basically combining the best exposed parts of each input image.

Seems to me that that is how every HDR creation tool, including PL, works.

Maybe I didn't explain the difference well: When you create a HDR, you re-combine the images according to their different exposures – so if you use a part of an input image which was overexposed by 2 steps, then you underexpose that part by 2 steps before it's added to the final image. In that way, you create an HDR image which has the same tones as the scene. enfuse doesn't even need to know about the exposure values and doesn't do any exposure adjustment, it just combines (blends) the original images.

Cheers

Burkhard.

[photoken]

Maybe I didn't explain the difference well: When you create a HDR, you re-combine the images according to their different exposures – so if you use a part of an input image which was overexposed by 2 steps, then you underexpose that part by 2 steps before it's added to the final image. In that way, you create an HDR image which has the same tones as the scene.

I don't think so, at least in PL's case, because the resulting image doesn't retain the same tones -- that's why one has to boost both the saturation and the contrast of the final image.

Luminance HDR, however, does allow one to alter the EV value of the incoming images to affect its final result, but I'm not sure it's algorithm compensates for EV as you've described rather than just combining the images using the specified overall EV settings.

HDRMerge uses a somewhat different approach and aggressively chooses the highest non-clipped values for the final image on a pixel-by-pixel basis. It can do that because it's working directly with the RAW data and creating a RAW (DNG) final result -- no de-mosaicking is involved.

At any rate, I don't see much (if any) difference between the results that can be obtained with PL and with Enfuse.