piccure+ releases update

[Bart_van_der_Wolf]

The sharpened image is clearly much sharper, but suffers from a fatal flaw - the high contrast edges look dreadful, e.g. on the three women's hijabs (head scarves).

For images that don't have high contrast edges, or edges of the kind caused by using an extremely high aperture lens wide-open, Piccure+ can produce extremely impressive results. But for those images with the flaw, it's a royal pain in the butt to fix it in Photoshop, which may or may not be worth it.

I reported this problem long ago. I don't know if it can be fixed or not. I hope the developers can fix it!

Hi Damon,

Have you tried reducing the rendering control (to something closer) to zero, i.e. more emphasis on detail than contrast?

I've 'played' around a bit with Version 3.0 on some different files, and I like the potential ability of P+ to correct spacially variant Capture blur, and the fact that the blind deconvolution is of decent quality. However, it's the fully automatic detection with very little manual override capability that produces results that can be improved upon by other deconvolution solutions, or more extensive sharpening suites.

Cheers,
Bart

[Damon Lynch]

Hi Mark and Bart, regarding printing, I'm afraid I have no idea because due to currently being in Tajikistan I haven't been able to make any prints for several years.

Regarding image noise, yes there's plenty -- welcome to the world of pre-2016 era Canon sensors! The version on the left has luminance NR set to 9 (detail 50) in LR and shadows at +12.

If rendering is set to zero, the problem with high-contrast edges is exactly the same, i.e. it makes no difference what value it's set to.

Attached is roughly the same crop, but this time from LR with LR capture sharpening set to the "Faces" preset (rather than set to zero as with the version above).

Finally, for a more favorable view of Piccure, I've attached a crop of a 100% view with the same body and lens at 11mm, this time tripod mounted, ISO 100. Under controlled conditions like this, it really shines. (sorry about the file size but I felt exporting at 80% jpg was necessary to preserve the fine detail)

[Mark D Segal]

Thanks Damon - interesting - I was once upon a time gearing up some work from home on a major project in Tajikistan, but that particular arrangement fell through :-) , so I never got there.

OK, back to photos, thanks for taking the trouble doing that. It would be good to see your LR settings in the Detail panel. Looking at the two images side by side, I don't see a huge difference in respect of sharpness and noise - the Piccure is a bit "crisper" all round, but detail not necessarily better (however, the detail in the ceiling of the other photo is impressive). Also not much difference in noise. These kind of tests are helpful for evaluating the merits of a new piece of software.

[Damon Lynch]

You're welcome Mark. The LR settings are nothing special - just as I described, plus highlights pulled down which I forgot to mention.

Attached is a 100% crop of this image: https://www.flickr.com/photos/damonlynch/21084192720/in/album-72157605715069583/

This attached image is my final copy, i.e. after I've done all post-processing, including the use of Topaz Detail. Again the lens is at 11mm in this image, using a tripod of course (actually the camera was just off the floor, with the legs as flat as they'd go). For this kind of image in my opinion Piccure in combination with other tools does a very fine job and is worth the effort.

[Mark D Segal]

Thanks Damon. Good sample and nice job. At the risk of trying your patience, may I ask how this same image would fare in LR?

[Damon Lynch]

Thanks Damon. Good sample and nice job. At the risk of trying your patience, may I ask how this same image would fare in LR?

To be honest I can't show you (at least not without a lot of work) - my regular workflow these days is to write over the top of the intermediary 16 bit TIFF, which in the case is the result of using LR first, SNS-HDR second.  The reason I don't save the intermediary TIFFs is that I found I never use them should I need to revisit an old file. I never use them because the software I use to work on them always improves. For example in this image my guess is the very latest SNS-HDR would do a better job that the version I used to produce the image you see.

[Mark D Segal]

Understood Damon, thanks. Of course when putting that question I had in mind my very much more simplified workflow which is to remain in LR from ingestion to print unless I need to do something to a photo that LR simply can't handle because it doesn't have the tools or the tweaks within tools needed for the purpose. 99% of my files never leave the raw state and the history is permanently available as long as the LR catalog and XMPs are viable. It's a different approach. I have a bunch of plugins to LR that require conversion to a TIFF version, so I only deploy them when I expect real value-added. This approach, by the way, really forces one to drive LR's capabilities to the limits - kind of a stubborn mindset to stay with the mothership, which is instructive in its own right. Cheers.

[Damon Lynch]

I probably end up working with TIFFs much more than you do Mark, but I hear you loud and clear regarding the downsides. The loss of state is probably the biggest. I wish every step could be faithfully recorded in XMP, ready to be recreated at moment's notice with the same software, just like we can in LR. Alas!

[Mark D Segal]

And how do you fare with electricity to run the computers and accessories over there? When I was working in Nepal (on that very sector BTW) it was a recurrent nightmare.

[Damon Lynch]

And how do you fare with electricity to run the computers and accessories over there? When I was working in Nepal (on that very sector BTW) it was a recurrent nightmare.

Mostly good in the capital and larger cities. Much better than it used to be.

Sometimes I wonder about the possibility of photographic tours here -- because so few visitors come here, the chances of getting a unique shot no one has done before are considerably higher than in many other places. There's plenty of great scenery here, and some wildlife too, e.g. snow leopards. But the big problem, as always, is political instability and the possibility that a tour would be halted because of it.

[Jack Hogan]

Thank you both. If sharpening is the key incremental benefit I probably don't need this application, because I obtain very sharp photos from my Sony a6300+Zeiss lenses with moderate capture and output sharpening in LR. Some form of deconvolution is helpful to rescue detail when I mis-focus (i.e. pilot error), but I have never been satisfied with the ultimate quality (from a photographic perspective) of those rescue attempts (using demo versions of such software) and therefore usually trash photos that are out of focus when they should be in focus. Now that too may be pilot error - maybe I need to spend more time focusing on deconvolution technique (pun intended); but that will be for another day :-)

Mark,

I too am not impressed with the ability of deconvolution to restore out of focus images.

Hello Mark and Bill.  I believe that's because deconvolution cannot restore out of focus images: defocus and dof cannot be modeled accurately as convolution with a blurring kernel.  Of course one may apply any type of micro/mid/macro local contrast adjustments to attempt to improve the 'perceived sharpness' of such an image, such as USM or deconvolution of small/large radius.  But by then it's no longer image restoration, it's simply an adjustment to make the final product look more pleasing.

In images captured with excellent technique there are very few quantifiable sources of blur left (diffraction, AA, pixels, various lens aberrations).  Restoration by deconvolution works best when those are known (in type, if not quantity).

Also the a6300 does not appear to have an AA filter, its green channel hitting Nyquist around 0.25 cycles/pixel with an FE55mm at f/5.6.

Jack

[Mark D Segal]

Also the a6300 does not appear to have an AA filter, its green channel hitting Nyquist around 0.25 cycles/pixel with an FE55mm at f/5.6.

Jack

Interesting observation, because there is a great deal of uncertainty bred of unnecessary secrecy about this matter. In any case, with many of these cameras the most determinative variable for resolution of scene detail is the lens, not the sensor.

[Mark D Segal]

Mostly good in the capital and larger cities. Much better than it used to be.

Sometimes I wonder about the possibility of photographic tours here -- because so few visitors come here, the chances of getting a unique shot no one has done before are considerably higher than in many other places. There's plenty of great scenery here, and some wildlife too, e.g. snow leopards. But the big problem, as always, is political instability and the possibility that a tour would be halted because of it.

Interesting, but unfortunate, because primary concerns of a tour or workshop operator are reputational risk, liability and liability insurance. Anything that elevates those risks above a comfortable level would make it challenging to organize and market such events. Seems more like an environment for hardy souls with a high tolerance for risk and somewhat deep pockets to cover the uninsurable possibilities.

[Bart_van_der_Wolf]

Hello Mark and Bill.  I believe that's because deconvolution cannot restore out of focus images: defocus and dof cannot be modeled accurately as convolution with a blurring kernel.

Hi Jack,

Those are two different things and I can add some other detractors too. The point is that mathematically, convolution can be exactly reversed by a deconvolution, provided we can use high enough precision calculations. It's like dividing and multiplying, e.g. 1/3*3 = 1 , although some calculators insist on a different outcome due to lack of precision. Convolution is not much more than multiple multiplications and additions, and deconvolution does the inverse order of operations.

The real issue is finding the exactly correct Point Spread function (PSF) that caused the defocus , and then we also need to deal with noise. In practice we do not know the exact convolution PSF, so this then becomes a process of (user assisted) blind deconvolution with noise regularization. And even then, we will find that precision is not enough to fully restore all detail, although we can come a long way.

Attached I've added an example of a synthetic image, a noisy image would be harder to find a model for:
1. A sinusoidal star image,
2. An image convolved with a defocus kind of PSF,
3. The PSF,
4. The result after deconvolution.

I stopped at 10000 iterations of the special Deconvolution algorithm, because this small image already took some 47 minutes to complete, a full size image would take days.

t's not a perfect restoration, but it does achieve a significant restoration of what seemed to be lost. I'm just showing that there is more possible that acutance enhancement.

In images captured with excellent technique there are very few quantifiable sources of blur left (diffraction, AA, pixels, various lens aberrations).  Restoration by deconvolution works best when those are known (in type, if not quantity).

Yes, but then there will still be other factors, like the demosaicing of the sensor signal and of the noise.

Also the a6300 does not appear to have an AA filter, its green channel hitting Nyquist around 0.25 cycles/pixel with an FE55mm at f/5.6.

The presence of aliasing artifacts will again make it harder to restore the original image signal.

Cheers,
Bart

[Jack Hogan]

Interesting observation, because there is a great deal of uncertainty bred of unnecessary secrecy about this matter.

Well, this is where the uncertainty about the presence or absence of an AA in the a6300 ends



There is no AA signature in the MTF data of the a6300+FE55mm at f/5.6 (DPR studio captures), green channel in the vertical and horizontal directions shown.

Jack

[Mark D Segal]

Thanks Jack; I'm not familiar with the technique you are using to illustrate this point. Perhaps if you don't mind supplying a few lines of explanation about this test, and putting up a chart from a camera that is known to have an AA filter it would be of educational value to me and perhaps other readers here. Thanks in anticipation.

[Jack Hogan]

Thanks Jack; I'm not familiar with the technique you are using to illustrate this point. Perhaps if you don't mind supplying a few lines of explanation about this test, and putting up a chart from a camera that is known to have an AA filter it would be of educational value to me and perhaps other readers here. Thanks in anticipation.

Hi Mark,

That's a plot of the spatial frequency response (sharpness at varying levels of detail) of the shown camera and lens as captured.  If a current implementation of an AA is present its job is to make the curve tend towards zero around +/- 0.7 cycles/pixels, so that aliasing-inducing frequencies above Nyquist (0.5 c/p) are minimized.  You can see the difference an AA makes in this well known duo (Canon 5DS with AA, 5DSR sans AA):



The earlier curve of the a6300 does not show the effect of an AA filter.

Jack

[Jack Hogan]

Hi Jack,

Those are two different things and I can add some other detractors too. The point is that mathematically, convolution can be exactly reversed by a deconvolution, provided we can use high enough precision calculations. It's like dividing and multiplying, e.g. 1/3*3 = 1 , although some calculators insist on a different outcome due to lack of precision. Convolution is not much more than multiple multiplications and additions, and deconvolution does the inverse order of operations.

The real issue is finding the exactly correct Point Spread function (PSF) that caused the defocus , and then we also need to deal with noise. In practice we do not know the exact convolution PSF, so this then becomes a process of (user assisted) blind deconvolution with noise regularization. And even then, we will find that precision is not enough to fully restore all detail, although we can come a long way.

Attached I've added an example of a synthetic image, a noisy image would be harder to find a model for:
1. A sinusoidal star image,
2. An image convolved with a defocus kind of PSF,
3. The PSF,
4. The result after deconvolution.

I stopped at 10000 iterations of the special Deconvolution algorithm, because this small image already took some 47 minutes to complete, a full size image would take days.

It's not a perfect restoration, but it does achieve a significant restoration of what seemed to be lost. I'm just showing that there is more possible that acutance enhancement.

Hi Bart that's a good demo and I agree with your comments on aliasing and noise.

However you have shown a simulation of convolution and deconvolution with a known PSF in 2D, a tractable problem.  On the other hand my understanding is that in practical photography situations defocus cannot be modeled as a convolution with a PSF because it cannot be known without having a precise 3D map of the image in the first place: in practice the scene's depth/phase information within the region of interest gets scrambled to the point of no return. So any attempts at recovering the original image need to make strongly simplifying assumptions that when applied almost never satisfy - as observed by Mark, and I am sure most anybody who has tried to recover anything other than very mild defocus with current technology.  Even then, my guess is that other than in ideal situations the improvement is generically pleasing but not accurate.

Also interesting is the opposite problem, trying to add focus blur to an image and make it look credible.  Check the background: credible?



Not to this critical peeper.  However, interestingly, that we can recover quite accurately.

Jack

[Mark D Segal]

Hi Mark,

That's a plot of the spatial frequency response (sharpness at varying levels of detail) of the shown camera and lens as captured.  If a current implementation of an AA is present its job is to make the curve tend towards zero around +/- 0.7 cycles/pixels, so that aliasing-inducing frequencies above Nyquist (0.5 c/p) are minimized.  You can see the difference an AA makes in this well known duo (Canon 5DS with AA, 5DSR sans AA):

The earlier curve of the a6300 does not show the effect of an AA filter.

Jack

Thanks very much Jack - very helpful. Now I should feel happier I bought an a6300?  :-)

[Jack Hogan]

Thanks very much Jack - very helpful. Now I should feel happier I bought an a6300?

Yes, unless you photograph a lot of man-made stuff like fabric and architecture.  In that case no