Reverse WB plus another way to sharpen . .

[xpatUSA]

This is a post about alternatives to the Conventional Way for WB and sharpening.

For WB, we are taught how the human eye compensates for the light and generally sees anything that is white as white, almost irrespective of the time of day. So we dutifully set our camera WB accordingly, most of the time.

The other day, I took a shot of my street. It was sunny, the sun was at about 30 degrees rising - so the WB was set to daylight. The shot was very bright and absolutely dripping with contrast. The shadows however were quite long and I wondered idly if it could be made to look a bit more like morning:



Somewhat exaggerated for the purpose of illustration, of course. How was it done?

In RawTherapee, I went to Custom White Balance and played with the temperature and tint sliders. The temperature ended up (non-intuitively) in the cold direction and the tint ended up toward red a little.

In Lab mode, I lowered the Chromaticity and Contrast for a more "early-morning" look.

Comments welcome.

Moving right along, I like sharp pictures, perhaps because my eyes are pretty blurry (old and myopic). I tend to favor de-convolution rather than the almost universal unsharp masking. Today I was looking at some recent shots enhanced by Piccure+. Review here: http://www.northlight-images.co.uk/r...cure_plus.html

Check out the dead grass and twigs (before and after Piccure+) in this shot:



Camera/lens is a Sigma SD14 and likely the Sigma 17-50mm f/2.8.

Piccure+ does not depend on lens profiles but corrects based solely on the image content. Quite a program; not free though.

[rpcrowe]

There are so many new plug-ins and stand-alone programs appearing, not to mention new versions of established editing programs, that it is absolutely mind boggling...

[xpatUSA]

There are so many new plug-ins and stand-alone programs appearing, not to mention new versions of established editing programs, that it is absolutely mind boggling...

Absolutely! As a Sigma owner, I am "fortunate" in having far fewer choices of new or even old stuff, due to what I call "the Gap". That's where, in a 'list of supported cameras' it goes "Samsung . . [the gap] . . Sony".

I tend to stick with stuff that I know and that does the job. Was dragged kicking and screaming away from Windows XP when my then current computer died completely and I had to buy a new one - but, fortunately, I found one with Win 7 Pro on it.

Sharpening is fascinating for it's infinite variety of both methods and apps, just as you said.

[dennybeall]

Nice to know about alternatives if I ever have to get rid of Photoshop.

[DanK]

The temperature ended up (non-intuitively) in the cold direction and the tint ended up toward red a little.

Not sure I know precisely what you did, but if I understand, this isn't counterintuitive. Or, at least, it is the same as my understanding of Lightroom. In LR, if you move the temperature slider to a higher Kelvin temperature (colder), the image looks warmer. In fact, the color band over the slider has yellow at the high end (50,000), and blue over the low end (2,000). When you slide the temperature slider to the right (higher numbers), you are not telling the software to make the image colder; you are telling it that the temperature of the image as shot was colder, so it needs to compensate more in the other direction, by warming it. It's doing what the camera does if, for example, you go out on a shady day and set the WB for shade (a higher Kelvin temperature). The camera will warm the image more, to compensate for this, than if you set the WB for daylight.

Or am I misunderstanding this?

[xpatUSA]

Not sure I know precisely what you did, but if I understand, this isn't counterintuitive. Or, at least, it is the same as my understanding of Lightroom. In LR, if you move the temperature slider to a higher Kelvin temperature (colder), the image looks warmer. In fact, the color band over the slider has yellow at the high end (50,000), and blue over the low end (2,000). When you slide the temperature slider to the right (higher numbers), you are not telling the software to make the image colder; you are telling it that the temperature of the image as shot was colder, so it needs to compensate more in the other direction, by warming it. It's doing what the camera does if, for example, you go out on a shady day and set the WB for shade (a higher Kelvin temperature). The camera will warm the image more, to compensate for this, than if you set the WB for daylight.

I don't use LR, nor PS, so can't comment about them.

I wrote "non-intuitively" just for the benefit of those less knowledgeable than our good selves. I can go back and take it out if you feel that it is misleading in some way . . .

[DanK]

I don't use LR, nor PS, so can't comment about them.

I wrote "non-intuitively" just for the benefit of those less knowledgeable than our good selves. I can go back and take it out if you feel that it is misleading in some way . . .

No, I don't think it is misleading. My choice of words wasn't great. I just meant to point out that sometimes users think that the color slider does the opposite of what it does do, at least in LR.

[NorthernFocus]

I wrote "non-intuitively" for the benefit of those less knowledgeable . . .

I agree that WB is (at least initially) counter intuitive. Setting a cooler temperature results in a warmer picture. Similar to exposure compensation. Setting a negative EC value results in a brighter picture. However, I must say from the title of the thread I was expecting some mind shaking revelation. I do think too many people get wrapped around the axle trying to "correct" WB. IMO "correct" WB is whatever value results in the photo looking like the actual scene or at least the scene as remembered (or desired as in the example). Obviously because of where I live I deal with a lot of WB adjustments due to overcast skies, snow in shade, etc. If fully "corrected" for WB many scenes do not look like what one sees when viewing them in person. When it is heavily overcast evergreen conifer trees look very much dark blue-green. In sunlight they trend to green containing much more yellow. If a photo is "corrected" all the way to the "daylight" shade of green but the sky is overcast it simply doesn't look natural to anyone who has ever viewed a similar scene in person.

Regarding the Puccure+ sharpening software, I tested it a while back and was quite impressed. It definitely does the best, most natural looking sharpening of anything I've tried. I posted a thread here discussing same. I haven't yet purchased it but intend to do so. Was hoping for them to run a promotional discount

[xpatUSA]

I agree that WB is (at least initially) counter intuitive. Setting a cooler temperature results in a warmer picture. Similar to exposure compensation. Setting a negative EC value results in a brighter picture. However, I must say from the title of the thread I was expecting some mind shaking revelation.

Sorry to disappoint the expectation. The main thrust was that of using the WB sliders in the opposite direction to the normal corrective action. That may be old hat for many but it was new to me.

I do think too many people get wrapped around the axle trying to "correct" WB. IMO "correct" WB is whatever value results in the photo looking like the actual scene or at least the scene as remembered (or desired as in the example). Obviously because of where I live I deal with a lot of WB adjustments due to overcast skies, snow in shade, etc. If fully "corrected" for WB many scenes do not look like what one sees when viewing them in person. When it is heavily overcast evergreen conifer trees look very much dark blue-green. In sunlight they trend to green containing much more yellow. If a photo is "corrected" all the way to the "daylight" shade of green but the sky is overcast it simply doesn't look natural to anyone who has ever viewed a similar scene in person.

Yes, a reasonable description of the pitfalls associated with WB selection/adjustment in the conventional direction.

Regarding the Piccure+ sharpening software, I tested it a while back and was quite impressed. It definitely does the best, most natural looking sharpening of anything I've tried. I posted a thread here discussing same. I haven't yet purchased it but intend to do so. Was hoping for them to run a promotional discount

Thanks for the link. I bought it yesterday, using it stand-alone (I'm Adobe-free). In RawTherapee, I'm tending more and more to ignore their two global sharpening methods (USM and RL de-convolution) and I feel that Piccure+ will likely work for that final "sharpen" to suit the output medium.

[Jimr1961]

Hmmm, as a Rawtherapee user, I am now curious about what makes one de-convolution more suitable than the other. Are they both based on RL algorithm with slight modification? Are default iterations different?

I too see better results with RL De-convolution over unsharp mask for raw files, but I haven't played with increasing iterations and don't have Piccure+.

[xpatUSA]

Hmmm, as a Rawtherapee user, I am now curious about what makes one de-convolution more suitable than the other. Are they both based on RL algorithm with slight modification? Are default iterations different?

Hello Jim,

With RawTherapee's RL method my understanding is that, for best results, you should really know the blur radius (more properly, the point spread function) that you are trying to de-convolve. I further understand that RawTherapee assumes the blur to be Gaussian which may not always be the case.

On the other hand, Piccure+ appears to be using what is called "blind de-convolution", meaning that the blur form/size is unknown. Therefore, I deduce that it's algorithm analyzes the image content somehow and applies what it calculates to be parameters appropriate to that content.

If you google <(adaptive OR blind) deconvolution> you'll get a ton of links for further study . . .

I'll guess that the algorithms are different, and the number of iterations appears to set by a 'Lens Quality' slider.

I too see better results with RL De-convolution over unsharp mask for raw files, but I haven't played with increasing iterations and don't have Piccure+.

I often (but not always) end up with 0.57 radius and about 63 amount. Generally, I leave the iterations at default.

[ajohnw]

Blind deconvolution stems from attempts to refocus images. Out of curiosity I use that style of sharpening on the unsharp mask version but to be honest I wasn't convinced that it was a good example of unsharp mask sharpening.



John
-

[xpatUSA]

Blind deconvolution stems from attempts to refocus images. Out of curiosity, I [used] that style of sharpening on the unsharp mask version but, to be honest, I wasn't convinced that it was a good example of unsharp mask sharpening.



John
-

There was no unsharp masking applied to either image, left or right, I hardly ever use USM. Therefore, there was no "unsharp mask version". In other words it wasn't intended to be a comparison between USM and blind de-convolution, it was supposed to be a before and after, sorry to have misled y'all.

I've edited the OP to clarify.

[ajohnw]

I think RawTherapee has a big draw back on sharpening. Reduced images always need sharpening to suite them and it's not possible to decide on the degree needed in RT so I always save and do that with something else however most of the time it could be reloaded into RT and done at 100% of it's new full size after reduction. The GIMP could also be used but it's degree numbers etc are totally different. The other problem with RT is the amount that can be applied to a full sized image. It's usually plenty to tidy up a typical image but if it's being used to fix problems - wind movement with still a chance of recovering a shot the rad just doesn't go big enough.

As to using just unsharp mask on Ted's shot as it was posted I selected all but the borders and used a fairly heavy degree with a rad of 2 pixels. I generally do use 2 pixel rad on finally reduced shots.



That particular shot can be "improved" further by messing with contrast a bit. I used a bit too much of a certain style of tone mapping but RT's lab etc contrast sliders would do the same. Or contrast by detail level. Maybe both. Their tone mapping always brightens for me which can make it hard to use. The flunky colour option does work well if that is what is needed.

Overdone a touch to show the change. More or less the same selection.



I understand that Adobe products automatically apply a degree of sharpening when shots are reduced. I've tried to get RT to include the reduction in the processing pipe but they wont or it can't be fitted in none destructively so just do the reduction when the image is actually saved. I just load it up in another application Trouble with doing that is they usually get changed in other ways after the final sharpening.

John
-

[xpatUSA]

Reduced images always need sharpening to suite them . . . .

John
-

Off topic, I have never understood why people think that down-sampling images softens them. My experience shows the opposite to be true. The well-respected Bart van der Wolf explains:

"You can see that all standard resampling methods in Photoshop^® CS fail to avoid aliasing artifacts when downsizing."

http://kronometric.org/phot/processi...%20methods.htm

My own tests, repeated today, illustrate the increase in sharpening and in contrast of a test target:



First here are the edge responses for the target at full size 1593px wide (left side graph), then reduced using Lanczos3 to 500px wide (middle graph) and also 1000px wide (right side graph).



The pronounced overshoot for 500px wide indicates severe oversharpening even though Lanczos3 is the same algorithm as is used by RawTherapee. Also check out the edge widths 10-90% in pixels.

And now for the MTF:



Here the tale of local/micro contrast is told. Basically the greater the area under the curve below 0.5 cycles/pixel (Nyquist) the more the contrast. It is clear to me that the 500px image has much more than the original image. Also notice the MTF at Nyquist is about three times that for the original - a sure recipe for visible aliasing. It is of course quite common for an image that has no aliasing full-size to develop some during re-sampling to a lower size.

van der Wolf clearly recommends a little Gaussian smoothing before downsizing in order avoid artifacts caused by the re-sampling.

[ajohnw]

Personally Ted I feel that the unsharp mask I used on the unsharpened 1/2 of your shot demonstrates why some is needed no matter what down sizing technique is used. I don't think it has anything to do with aliasing. It more a case of "restoring" detail losses that are a result of the process.

This is a pretty typical view of sharpening.

https://helpx.adobe.com/photoshop-el...harpening.html

John
-

[xpatUSA]

Personally Ted I feel that the unsharp mask I used on the unsharpened 1/2 of your shot demonstrates why some is needed no matter what down sizing technique is used.

just to be clear, John, the posted image halves are 100% crops, so mentions of "down-sizing technique" are not relevant to those images.

I don't think it has anything to do with aliasing.

Oh dear, my previous post was clearly a waste of time or perhaps just not understood? I'm guessing the latter.

It's more a case of "restoring" detail losses that are a result of the process.

A bit too vague to respond to, sorry.

This is a pretty typical view of sharpening.

https://helpx.adobe.com/photoshop-el...harpening.html

John
-

Thank you for the link to 'Sharpening 101' a la Adobe but I don't use any Adobe products and it told me nothing new about the "conventional way" - sorry again.