Pages: [1] 2   Go Down

Author Topic: A critical look at demosaicing, capture sharpening and noise rendering  (Read 3174 times)

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446

I've been studying the way detail is rendered on a pixel level by different demosaicing algorithms and the various artifacts that each method produces, and how each renders noise grain textures. This in turn is amplified in different ways by different capture sharpening tools. In particular, almost all the sharpened examples I provide below are using Focus Magic, as this is the tool that is highly recommended by someone I highly respect, Bart van der Wolf.

Before someone comes along to say that I won't ever see any of these issues in print, for the record I do see them in prints. I have a bunch of images from years ago that are around 10-20MP, before I learned to stitch to get more native resolution. I have made average to medium-large sized prints (16x20", 20x25", 30x40") with them. I am interested to perfect these images on a pixel level so prints of those sizes can be made that will be largely artifact free.

Some time ago I transitioned from ACR to Raw Therapee to obtain the amazing AMaZE demosaicing. I found that on some of my images, where there are contrasting colors (yellow leaves against blue sky), I get significant mazing artifacts with ACR's demosaicing which I saw in my prints and that annoyed me. Textured surfaces like granite or tree bark or even skin texture in portraits do not reveal the deficiencies of the ACR demosaicing, and I had been quite happy in the beginning with ACR on more forgiving subjects such as these. Being a landscape photographer however, it is hard to avoid pictures where foilage might be set against a blue sky!

I have since noticed some disturbing artifacts with AMaZE demosaicing too - while it produces the best overall rendition of fine detail that has minimal mazing and zipper artifacts, some weird patterns form along diagonal edges of relatively high contrast. Note that you will not see these artifacts in edges that are vertical or horizontal or almost that. They are most obvious in diagonal edges (see first attached example, red arrows point to where one might see these artifacts). It is a sort of stair-stepped ringing artifact, with lines of varying contrast at exactly 45 degrees. I do not mean the different kind of digital jaggies, which is what happens with DCB demosaicing for example that is more coarse (you get 2x2 pixel blocks of steps rather than 1 pixel steps with AMaZE, which is the finest possible step). It's a single pixel wide line of darker dots, always a perfect 45 degree slant, and these lines often appear next to each other and overlap, similar to the kind of ringing artifacts one might see with Lanczos 8 interpolation. I wonder if the anti-zipper smoothing algorithm of AMaZE is the cause of this kind of artifact, making zipper artifacts - alternating dark pixels at a 45 degree angle - smooth perhaps means joining them up. Another artifact that is observed is a dark single pixel tends to appear in the corners of edges, like on the tip of a leaf. They are peppered throughout the provided example.

When I capture sharpen with Focus Magic, it enhances the stair-stepping to a very high degree, making edges look to have strong ringing artifacts. The black-dots artifact is also enhanced to a large degree. Using the blend-if sliders to tame the dark halo reduces the effect, but does not eliminate it as it is present already. I've noticed that using Topaz Detail's Deblur function, the ringing artifacts are hidden to a very large degree, except on some really problematic edges. I wonder if Focus Magic is creating these ringing artifacts - but a sanity check on a DCB demosaic version shows no such 45 degree diagonal lines. I also noticed that the thin-ness of edges, if there is indeed such a way of describing how crisp an edge is, is much finer than what I can achieve with even radius 1 in Focus Magic. The result is that edges appear to be more crisp, though at lower percentage viewing on monitors, the fatter halos with Focus Magic gives the sensation of a 'sharper' image, but when upsampled ~400% - 800% for printing larger and to serve 600ppi data to my Canon printer, Topaz Detail's Deblur (or Infocus) produces a visible cleaner edge which lends itself to more (better) output sharpening.

I was interested to see how good the latest Capture One V10 renders detail and grain pattern, as Bart has mentioned several times that the demosaicing has been significantly improved (since Ver 9 I believe). Indeed it has, but unfortunately in my comparisons I am not convinced that it is better overall than AMaZE. There is still quite some significant zipper artifacts (see second attached crop example), and the contrast of the zipper artifact pixels is quite high, so when sharpened, looks even worse than ACR in some regions (hunt for them in the full resolution tiff!). But it must be noted that ACR's demosaicing is significantly worse for mazing and zipper artifacts overall, with lots of false color, and has a weird soft look to the detail perhaps owing to what looks like noise filtering. Capture One and Raw Therapee (with AMaZE, DCB, LMMSE and AHD) all exhibit much more crisp rendition of detail with varying levels of mazing and zipper artifacts, but the grain structure is correspondingly crisper but harsher, requiring some luminance noise reduction to reduce it to ACR levels. I notice that Capture One V10's grain structure has the occasional relatively very light and very dark pixel that appears sporadically, and some correlations in the noise pattern occurs (with zero sharpening of any sort). I also see that AHD creates some kind of weird hash pattern (almost like a JPEG artifact) in the grain, which can be seen also with DCB and LMMSE, and AMaZE even, but it is less so in the latter. Capture One exhibits a small amount of the hashing pattern, visible in the second attachment where it is enhanced by the Focus Magic sharpening.

I am providing here a few full resolution examples to show what I am referring to, including links to the original raw test image. These are layered tiffs (warning: huge) containing the full resolution renditions, converted with different demosaicing from ACR 9.1.1, Raw Therapee and Capture One V10, and then a copy of each with Focus Magic sharpening applied. The layer names should be self-explanatory. The bottom set of layers, without the 'FM 1 300%' suffix in the layer name, is just converted from raw, with the white balance and exposures matched as best as I can, with zero curves, sharpening, noise reduction and all other adjustments zeroed out. You can try your own editing on the sample raw file if you wish, if you don't trust my layered tiff and wish to come to your own conclusions.

Download Links:
Raw image sample. Canon 5D II capture. For personal use only please.

https://drive.google.com/open?id=0B5AXKSbQEPuFODBkVjBwLWlwRFk


Layered Tiff (1.3GB) This is the above raw image converted. I included the AMaZE rendition sharpened with Topaz Detail's Deblur as a comparison to the Focus Magic Sharpened layer. Toggle the layer visibility for best visual comparison (easier to see than the side-by-side example crops below) Note that I used Focus Magic at 300% amount to enhance the artifacts of the sharpening process, and I NEVER use such a high amount ever for capture sharpening, but the artifacts are still there, just less visible/amplified. That is, the width of the halos, and the amount of ringing, is the same but with less contrast only at lower amount percentages. Pan about at 300% view or even more - there is a fascinating minefield of artifacts to look at. Flip between the unsharpened and no noise reduction originals (bottom layers) and the sharpened layers. It's interesting to see how each demosaicing algorithm renders detail on a real world image, rather than on some synthetic target. Some leaves show significant mazing around them in the blue sky region, some leaves show significant zipper artifacts. Some don't show any. Complex results.

https://drive.google.com/open?id=0B5AXKSbQEPuFU1hNMnhxNXQ5Ujg


Layered Tiff (2GB) This is a different example, a frame from a Sony A7R II, underexposed and with extremely low contrast and shot at ISO 100 which makes it a good candidate to look at fine grained noise when the data is stretched quite a lot with a curves adjustment layer. AMaZE renders the grain of the image quite harshly compared to ACR, with random extra light and dark pixels peppered about. ACR has perhaps the best noise grain pattern, which is not just softer but the intensity of the salt and pepper noise is leveled out to around the average deviation. It looks like ACR has some noise filtering going on in the demosaicing, despite the noise reduction settings zeroed out. This example has large flat gray regions of almost no detail to allow a study of the rendition of grain patterns by the various demosaicing algorithms, but also some contrasty detail in the bottom corner to look at artifacts on high contrast edges too. Not as good as above example for looking at edge artifacts however.

https://drive.google.com/open?id=0B5AXKSbQEPuFbGJjWVBVZXNIdUU



It would be interesting to hear if anyone has found ways to mitigate such artifacts, whist maintaining the highest possible amount of detail and crisp edges that looks good. Bart, if you happen to see this and have the time to comment, it would be most appreciated. It is always enlightening to read your thoughts on achieving high quality results. I find it interesting that I prefer Deblur's rendering of edges over Focus Magic, which I have long held as the better capture sharpening tool because you said so on several occasions. I also find it weird that I cannot find information detailing the issue of diagonal patterns with AMaZE demosaicing.

I suppose overall I find AMaZE to give the best rendition of fine detail, but it needs to be coupled with Topaz Detail's Deblur or Topaz Infocus rather than Focus Magic to avoid making the majority of the 45 degree diagonal ringing lines visible, and one can use the blend-if sliders to tame the peppering of dark almost-black pixels on certain edges. I just wish AMaZE will not generate the disgusting diagonal lines!
« Last Edit: February 06, 2017, 12:45:12 PM by samueljohnchia »
Logged

MichaelEzra

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 858
    • http://www.michaelezra.com
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #1 on: February 06, 2017, 05:54:18 PM »

I would highly recommend Smart Sharpening using Photoshop CC (2015-17).
The deconvolution methods used here render noticeably higher resolution detail.

With regard to demosaicing. there is no silver bullet. I sometimes blend results of RawTherapee/AMAZE and latest ACR with hand painted masks to get the absolute best rendition.

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #2 on: February 06, 2017, 07:56:23 PM »

Thanks Michael. I currently do not have Photoshop CC but I can download a trial to test and see. But if Smart Sharpen is so good, I wonder why Bart does not recommend it over Focus Magic. Perhaps the current implementation is now better at not enhancing noise in images.

The deconvolution methods used here render noticeably higher resolution detail.

I'm not sure what you mean. The methods you refer to are Focus Magic and Topaz's deconvolution methods? Or Smart Sharpen? Higher detail than?

I am well aware that there is no perfection to be found with demosaicing. But perhaps there may be further improvements to be had by combining the strengths of various methods, or by filtering out the artifacts with other methods. I tried eliminating the black pixel peppering with Topaz Denoise in the AMaZE version, but it is not effective enough. Perhaps wavelets in Raw Therapee can filter out this kind of noise (haven't tried yet) but it is quite low contrast before any sharpening is applied to amplify the issues, so I wonder how effective that can be too. Also, surely there must be better ways of smoothing zipper artifacts without chaining them up into ringing diagonal halos. I have yet to make up my mind which kind of artifact looks worse in print, assuming they are visible to the same degree. Mazing is of course the worst. But zipper dotted edges or fake digital-looking coalesced-into-45-degree-lines? I'm not so sure.

I have blended by hand different interpolated results many times before. It's many hours of hard, boring looking, a real time eater. There are thousands of artifacts that need to be fixed in just that one 21MP example. It would be nice to have a method that takes care of 50%-80% of the most problematic artifacts in one fell swoop, automatically. Hence my first decision to abandon ACR for Raw Therapee's demosaicing.
« Last Edit: February 06, 2017, 08:02:56 PM by samueljohnchia »
Logged

scyth

  • Sr. Member
  • ****
  • Online Online
  • Posts: 532
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #3 on: February 06, 2017, 09:19:30 PM »

Canon 5D II capture.

I am so sorry - but I'd rather pay $ to buy 5Ds/r (or even A7R2) than sweat time & effort demosaicking that last "1%" (or "5%") from 5DII raws ...
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #4 on: February 06, 2017, 10:11:22 PM »

I am so sorry - but I'd rather pay $ to buy 5Ds/r (or even A7R2) than sweat time & effort demosaicking that last "1%" (or "5%") from 5DII raws ...

The demosaicing artifacts are present even in 5Ds and Sony A7R II raws. And I do own the Sony, I paid the $$
Logged

MichaelEzra

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 858
    • http://www.michaelezra.com
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #5 on: February 07, 2017, 02:01:11 PM »

Here is a quick illustration of CC's smart sharpen and the "basic" RT sharpen (threshold-controlled USM).

To fight the pepper and salt noise in RT, try Impulse Noise Reduction.

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #6 on: February 07, 2017, 11:13:13 PM »

Here is a quick illustration of CC's smart sharpen and the "basic" RT sharpen (threshold-controlled USM).

To fight the pepper and salt noise in RT, try Impulse Noise Reduction.

Thank you. I've downloaded PS CC and given Smart Sharpen a close look, and also compared it to CS6. CC's smart sharpening is different, and better implemented than CS6. It avoids halos and ringing artifacts better than CS6, and the effect is less strong for the same settings (allowing finer adjustment). But it is significantly worse than Topaz Detail's Deblur/Infocus in that it sharpens the mazing, zipper and diagonal line artifacts just as badly as Focus Magic while Topaz mostly avoids doing so. CC Smart Sharpen is also worse at avoiding sharpening noise compared to Focus Magic or Topaz Detail's Deblur/Infocus. When using Smart Sharpen's noise reduction, the sharpening of the noise is suppressed, but the grain pattern in smooth areas (blue sky in my example) takes on a hashing pattern, which is extremely ugly.

As it stands, Bart's recommendations for using Focus Magic or Topaz Infocus/Detail's Deblur for capture sharpening over other methods remains sound to me.
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #7 on: February 12, 2017, 12:49:16 AM »

Does anyone have any explanation for the diagonal line artifacts that is produced by using AMaZE demosaicing in Raw Therapee? There is often a lot more discussion about false color and aliasing issues, even mazing and zipper artifacts, but I have not found any documentation about this behavior/character in demosaicing. Most people simply declare AMaZE as the overall best quality demosaicing available today.

In this post I include one more comparison between AMaZE and DCB, which is pretty close to Lightroom/ACR's result in this section. Sharpened with Focus Magic, radius 1, amount 175% which makes the artifacts more noticeable. 400% magnification
Logged

sandymc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 331
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #8 on: February 12, 2017, 05:07:06 AM »

AMaZE attempts to do something the authors call "diagonal interpolation correction" after the main interpolation. You would probably need to try with that switched off to be sure, but probably AMaZE is getting a little too enthusiastic with the correction.

Sandy
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #9 on: February 14, 2017, 04:34:42 AM »

Thanks Sandy. I've never heard of that before. It sounds like it could be the cause of what I am seeing. However I don't see any way of switching it off in Raw Therapee. Maybe I should write to the authors. Do you know who might be good to contact about this?
Logged

Tony Jay

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2840
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #10 on: February 14, 2017, 04:51:59 AM »

The demosaicing artefacts that you refer to cannot be escaped.
Simply put, any colours in a scene adjacent to one another that are opposite one another on a colour wheel will result in a white (or near enough) line separating those colours in a demosaiced raw image.

A common result of a blue sky adjacent to yellow foliage or geology is an annoying white line outlining the foliage or rocks. The reason is that part of the demosaicing process requires a sampling of nearby pixels in order to decide what the colour of any particular pixel should be. If blue and yellow abut each other then in RGB colour terms mixing blue and yellow gives white (or close). This is because in RGB colour theory colours are additive.
However the phenomenon is not limited to yellow/blue and can occur with any two colours that are opposite on a colour wheel.

The first person who convincingly explained this phenomenon to me was none other than our own (LuLa) Jeff Schewe in the Camera to Print and Screen series and subsequently also in his book "The Digital Negative."
Depending on how intrusive this is in an image (sometimes it almost invisible since the sky can sometimes be almost white close to the horizon) it may require layers and masks in Photoshop or some other equally capable pixel editor to fix.

I will leave others to comment on the sharpening aspects raised.

Tony Jay
« Last Edit: February 14, 2017, 05:00:29 AM by Tony Jay »
Logged

sandymc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 331
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #11 on: February 14, 2017, 05:19:27 AM »

Thanks Sandy. I've never heard of that before. It sounds like it could be the cause of what I am seeing. However I don't see any way of switching it off in Raw Therapee. Maybe I should write to the authors. Do you know who might be good to contact about this?

Probably the best way is via their forum. In my experience, they're very responsive.

Sandy
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #12 on: February 14, 2017, 09:57:11 AM »

The demosaicing artefacts that you refer to cannot be escaped.
Simply put, any colours in a scene adjacent to one another that are opposite one another on a colour wheel will result in a white (or near enough) line separating those colours in a demosaiced raw image.

Hi Tony, thanks for your input.

In my examples, I show that the demosaicing artifacts with one method of demosaicing can be overcome by a better method, but other artifacts of a completely different nature result. What I asked was if the good qualities of different methods can be combined to provide a superior result.

It has absolutely nothing to do with white lines appearing to separate opposite colors on a high contrast edge. I have no idea what you are talking about, relating this phenomenon to the examples I provided, which do not exhibit any such flaw. I have read about this issue in 'The Digital Negative', but I've never had this problem myself with Raw Therapee's AMaZE demosaicing, nor with ACR/Lightroom. The artifacts I detailed in my OP are of a completely different kind.
« Last Edit: February 14, 2017, 10:50:49 AM by samueljohnchia »
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #13 on: February 14, 2017, 10:00:06 AM »

Probably the best way is via their forum. In my experience, they're very responsive.

Sandy

Thank you, I will definitely post and share this on their forum. Hopefully the result will be improvements to the AMaZE demosaicing, which is already the best in all respects, apart from this weird diagonal interpolation artefact.
Logged

BartvanderWolf

  • Sr. Member
  • ****
  • Online Online
  • Posts: 6952
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #14 on: February 14, 2017, 11:09:32 AM »

Thank you, I will definitely post and share this on their forum. Hopefully the result will be improvements to the AMaZE demosaicing, which is already the best in all respects, apart from this weird diagonal interpolation artefact.

Hi Sandy (sorry I intended to reply to) Samuel,

All these algorithms are attempting to recreate and quantify the original scene signals from a sparsely sampled grid. It will never be perfect because some of the required info is missing. It is remarkable though, how much can be done with so little information. The better algorithms, like AMaZE and CaptureOne do a better than average job on most images. To improve, one could provide better input, such as captures with sensor shift and multiple exposures, or 'trained systems' that do pattern recognition.

I once read a comment by Emil Martinec, that the code in RawTherapee is not the full implementation of his code that he prototyped in MathLab, for speed reasons (if I remember correctly). So there might be a tiny bit more untapped potential in AMaZE.

The minor artifacts that you've shown, would probably be mitigated a bit by not sharpening the source conversion, but by using a Blend-if layer (e.g. like attached as a starting point) for sharpening. That would hold sharpening back a bit if edge contrast is already very high (sharp enough).

Cheers,
Bart

P.S. Other than Photoshop, Affinity Photo also offers such control with blending of a Sharpening layer that depends on lower layers in the stack, and it even allows to modify the gamma of the blending layer and allows a non-linear transition, example attached.
« Last Edit: February 17, 2017, 08:07:55 PM by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

sandymc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 331
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #15 on: February 14, 2017, 02:15:47 PM »

Hi Sandy,

For clarity, the original question, and report of the artifact(s) was Samuel's. I just commented that the nature of the AMaZE algorithm might explain some of what he was seeing.

Sandy
Logged

BartvanderWolf

  • Sr. Member
  • ****
  • Online Online
  • Posts: 6952
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #16 on: February 14, 2017, 02:34:54 PM »

For clarity, the original question, and report of the artifact(s) was Samuel's. I just commented that the nature of the AMaZE algorithm might explain some of what he was seeing.

Sorry Sandy, my mistake.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Tony Jay

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2840
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #17 on: February 14, 2017, 04:10:04 PM »

Hi Tony, thanks for your input.

In my examples, I show that the demosaicing artifacts with one method of demosaicing can be overcome by a better method, but other artifacts of a completely different nature result. What I asked was if the good qualities of different methods can be combined to provide a superior result.

It has absolutely nothing to do with white lines appearing to separate opposite colors on a high contrast edge. I have no idea what you are talking about, relating this phenomenon to the examples I provided, which do not exhibit any such flaw. I have read about this issue in 'The Digital Negative', but I've never had this problem myself with Raw Therapee's AMaZE demosaicing, nor with ACR/Lightroom. The artifacts I detailed in my OP are of a completely different kind.
Fair enough, I was commenting about issues which tend to be present no matter which demosaicing algorithm is used. I did base my reply more on the written parts of the OP and not particularly from the two posted images.

Tony Jay
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 446
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #18 on: February 14, 2017, 06:11:34 PM »

Hi Sandy (sorry I intended to reply to) Samuel,

Hi Bart, so good that you can chime in!  :) Thank you.

Quote
All these algorithms are attempting to recreate and quantify the original scene signals from a sparsely sampled grid. It will never be perfect because some of the required info is missing. It is remarkable though, how much can be done with so little information. The better algorithms, like AMaZE and CaptureOne do a better than average job on most images. To improve, one could provide better input, such as captures with sensor shift and multiple exposures, or 'trained systems' that do pattern recognition.

Indeed these demosaicing algorithms are impressive, creating so much convincing detail with so little original information. I understand that perfection is not possible, but by looking at the strengths and weaknesses of each method, I wonder if it is possible to combine some of the strengths. In this case, AMaZE's demosaicing, but without the diagonal interpolation artifacts will be good enough for me. I am not bothered by false color artifacts (there is already a function in Raw Therapee that can help tame that), or by the noise pattern created by AMaZE.

What I found disappointing was that I found Capture One V10 to be significantly worse at avoiding zipper artifacts, which I thought it would be better, going by your high praise for it. It is pretty good at avoiding mazing artifacts, but I would not even consider it better than DCB overall, not to mention inferior rendition of the noise grain pattern.

Quote
I once read a comment by Emile Martinec, that the code in RawTherapee is not the full implementation of his code that he prototyped in MathLab, for speed reasons (if I remember correctly). So there might be a tiny bit more untapped potential in AMaZE.

Fascinating. So perhaps a "ultra high quality" implementation, at the cost of whatever-it-may-be, may be possible with even higher demosaicing quality.

Quote
The minor artifacts that you've shown, would probably be mitigated a bit by not sharpening the source conversion, but by using a Blend-if layer (e.g. like attached as a starting point) for sharpening. That would hold sharpening back a bit if edge contrast is already very high (sharp enough).

The diagonal line artifacts caused by AMaZE are so strong that they are visible even with no sharpening at all. I always use the Blend-if sliders in combination with any sharpening layers (just not in these tests, but easy to turn them on in my layered TIFF example to see the effect), but they are almost completely useless at mitigating the enhancement of the diagonal interpolation artifacts created by AMaZE, since these artifacts do not lie on the extreme ends of the tone scale.

In my later experiments, I've noticed that it is possible to generate much smaller width halos at low settings with Topaz Detail's Deblur/Infocus than with Focus Magic at it's smallest radius of 1. I suspect you might raise the issue of inadequate sharpening to counter the overall system blur and to restore the original signal, but I cannot help but notice how much better looking Deblur looks in my example, avoiding sharpening a lot of artifacts with zero user intervention. Photoshop CC's Smart Sharpen and Focus Magic enhances artifacts in almost similar ways (FM just avoids enhancing noise way better), which leads me to suspect if they employ the same method of deconvolution with similar blur kernel assumption, while Deblur uses a completely different method altogether which seems to target 'true' edges only, thus avoiding the diagonal line artifacts which lie, one to several pixels, away from the true edge. Pretty darn amazing. At 100% view however, because FM's sharpening extends quite symmetrically towards both ends of the tone scale, the dark edge contours are darker and give the illusion of greater texture (and therefore more apparent detail), than Deblur, but at the cost of fatter structures in the detail, something I deplore.

Affinity Photo has intriguing options for blending layers. I'm really glad to learn this. When I've got enough of Adobe, it great to know that there is a good substitute available.
« Last Edit: February 14, 2017, 06:19:31 PM by samueljohnchia »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 591
Re: A critical look at demosaicing, capture sharpening and noise rendering
« Reply #19 on: February 14, 2017, 07:32:15 PM »

Improved sharpening together with significant alias artifact reduction could be done by taking two exposures, one with a small aperture of F22.

The F22 image would have greatly reduced aliasing but poor resolution. But, because it is alias free, it would provide information which could determine the actual degree of aliasing amongst the different pixels in the reference image. This should produce much better sharpening and reduce alias artifacts, especially false color ones or pattern moire at the same time. Seems like a potentially useful tool useful for landscape photography amongst others. I keep expecting this to be productized. Perhaps it has been and I'm not aware of it.
Logged
Pages: [1] 2   Go Up