Pages: 1 ... 20 21 [22] 23 24   Go Down

Author Topic: A free high quality resampling tool for ImageMagick users  (Read 240560 times)

Dogway

  • Newbie
  • *
  • Offline Offline
  • Posts: 8
Re: A free high quality resampling tool for ImageMagick users
« Reply #420 on: March 24, 2015, 06:18:16 am »

PS: maybe I should just test but I guess -set filter:lobes won't work with -filter spline right?
No, it does not. Only (TTBOMK) with windowed Sinc and Jinc, and Lagrange (the closest thing to "arbitrary number of lobes (polynomial) splines" in ImageMagick).

Thank you, I just tested and Lagrange 3 lobes is pretty much CatRom, using 2 or 4 lobes is even sharper so I don't know. Anyways I'm very glad with RobidouxSharp results and will stick with that, I thought on adding some post sharpener ala Unsharp Mask in Photoshop, just testing around got similar values for those I use there, one could argue what could be better, if stressing a bit more on the "c" component or using a post-sharpener, but I like to respect kernel conventions and work from there.

I post what is IMHO the state of the art for downscaling (fill the rest of settings according).
You should only tweak the 1.2 amount value of unsharp, set to 0 to disable.

-define filter:b=0.2620 -define filter:c=0.3690 -filter Cubic -distort resize 25% -unsharp 0x0.3+1.2+0.0
« Last Edit: March 24, 2015, 06:20:01 am by Dogway »
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #421 on: March 24, 2015, 06:38:13 am »

...
-define filter:b=0.2620 -define filter:c=0.3690 -filter Cubic -distort resize 25% -unsharp 0x0.3+1.2+0.0
You can get the same (more precise actually:.2620 and .3690 are rounded values; internally, IM uses full precision versions of the corresponding algebraic numbers) with
-filter RobidouxSharp -distort resize 25% -unsharp 0x0.3+1.2+0.0
Logged

Dogway

  • Newbie
  • *
  • Offline Offline
  • Posts: 8
Re: A free high quality resampling tool for ImageMagick users
« Reply #422 on: March 24, 2015, 06:44:12 am »

I forgot to ask, is it ok to unsharp in linear light (or high bit depth), or do I need to do after that? Sorry with the newbie questions...
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #423 on: March 24, 2015, 06:45:51 am »

Generally, you should get better results if you resample in linear light, but post-sharpen in gamma light.
« Last Edit: March 24, 2015, 11:02:49 am by NicolasRobidoux »
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: A free high quality resampling tool for ImageMagick users
« Reply #424 on: March 24, 2015, 11:01:08 am »

I forgot to ask, is it ok to unsharp in linear light (or high bit depth), or do I need to do after that? Sorry with the newbie questions...

Hi,

Deconvolution is most predictable in linear light, because it will affect shadows the same as highlights. But that is more about restoration of the original signal's resolution than traditional USM 'sharpening'. Unsharp masking is more about boosting acutance, edge contrast, and that can be done quite well on gamma adjusted images.

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

Dogway

  • Newbie
  • *
  • Offline Offline
  • Posts: 8
Re: A free high quality resampling tool for ImageMagick users
« Reply #425 on: March 24, 2015, 08:21:28 pm »

Thanks. I guess I can just use you employed convolution method "DoG" for RobidouxSharp, changing some settings like strength or radius so there's not much dark halos. Maybe even with your blending approach, although I don't know if there are benefits on that regard when gamma encoded resizes.

I did a brief test with the monkey pic. Took some 2 or 3 meters distance and toggled between original, linear and gamma encoded versions to fit screen (with Photoshop preview upscaling). The linear version was pretty much the same as original even at a distance, while the gamma encoded was darker, so I guess that yeah, if you were from an object at a larger distance you would see things like in linear light resizes.
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 798
    • Hikes -more than strolls- with my dog
Re: A free high quality resampling tool for ImageMagick users
« Reply #426 on: March 27, 2015, 02:14:52 pm »

You can get the same (more precise actually:.2620 and .3690 are rounded values; internally, IM uses full precision versions of the corresponding algebraic numbers) with
-filter RobidouxSharp -distort resize 25% -unsharp 0x0.3+1.2+0.0

Thanks for this, Nicolas and Dogway.

Just to throw in a slightly different perspective, I decided to measure MTF curves of a downsized ISO100 D800e slanted edge using the different methods discussed in the latter part of this thread.  The original image is a 300x400 pixel crop directly off DSC_6483.NEF (I downloaded it from dpr and so can you if you so wish) saved as a tiff with absolutely zero processing other than CFA normalization.  I think it's as close as one can get to the actual spatial resolution information captured by the camera.  You can download it from here and it looks like this



That beautiful edge was first measured with MTF Mapper - then downsized 4:1 using Photoshop's standard methods and re-measured.  These are the Spatial Frequency Response (MTF) curves of the original image indicated by a dashed line and after downsizing to 25% of its original size by CS5 nearest neighbour, bilinear and plain vanilla bicubic.



The original is soaring at those lofty heights because one of its pixels corresponds to four of the others'.  Nearest neighbour tracks it well, which is good at the lower frequencies, not so good beyond Nyquist because the ideal SFR/MTF curve for a perfect downsizing would track the original's up to Nyquist - but show little or no energy above 0.5 cycles/pixel, the point at which aliasing and moirè start rearing their ugly heads. Bilinear and bicubic show different amounts of attenuation, with bicubic looking like the best compromise.  The blue curve below shows the upper boundaries of what I think an ideal MTF curve should look like after perfect downsizing only (no extra sharpening), with bicubic for reference as the dotted line.



Next I plotted the MTF curves from three downsizing algorithms discussed in this thread ('D' option in the script): 'nodownsharp' quadratic, 'downsample' V1.22 and RobidouxSharp.  None were sharpened after downsizing, with 'downsample' set at a minimalist 1% DoG.  Quadratic and downsample 1% fare much worse than our benchmark bicubic while RobidouxSharp does a good job against it at the cost of a little extra energy beyond Nyqyist.



Lastly I decided to give Bart his due and show 'downsample' V1.22 as it was meant to be, with full 100% DoGs.  The following plot shows how the V 1.22 'downsample' blend works: downsizing first, followed by restoration:



The 1% dotted red curve is the heavily attenuated result of its downsizing component.  With 100% 'sharpening' the DoGs component (solid red line) attempts to make up for the loss of energy in the desirable frequency range by amplifying the attenuated curve (and everything else including unwanted noise) back to where it thinks it should be.  I can't help but wish that both the attenuation and the amplification not be so drastic, although it does do well below Nyquist at the cost of an overshoot in the lower frequencies and lingering on in the higher ones.  

Thoughts?

Cheers,
Jack
« Last Edit: March 28, 2015, 06:06:06 am by Jack Hogan »
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #427 on: March 28, 2015, 04:29:17 am »

Roughly, RobidouxSharp is the "sharpest reasonable" cubic kernel when used with EWA. (Although some people like the look of EWA Catmull-Rom a.k.a. CatRom, which is even sharper, it is my opinion that this filter gives results which drift too far from "faithfulness". In my book, RobidouxSharp marks the end of "generally usable for EWA".)
EWA Robidoux is a good match for the Nyquist content of gently demosaicked Bayer source.
Even though EWA Robidoux is very popular for downsampling, there is some justification for using something slightly sharper, but not as sharp as RobidouxSharp. EWA Mitchell is a good start.
You could also directly choose the C value using "-define filter:c=YOUR_CHOSEN_VALUE -distort Resize". TTBOMK, you don't need to specify that you want a Cubic filter because the default (Robidoux) is one. In other words, the above should give the same as "-define filter:c=YOUR_CHOSEN_VALUE -filter Cubic -distort Resize". You can also set the "b" value instead of the "c" value, in which case the usable range is 1 (Spline = very soft) to 0 (CatRom = very sharp). "B" is amount of "blurriness" (easy enough to remember).
Terminology: EWA = Elliptical Weighted Averaging = what you get when you use "-distort Resize" instead of "-resize" with ImageMagick. It's also the basis of the "Jinc" methods of madVR. EWA is actually slowly spreading. It is also now available for Blender and a few other FLOSS image and video processing programs.)
P.S. Some people like the look of BC-cubics that are not Keys (for which 2B+C=1). In particular, methods with B+C=1 are popular. There is a strong theoretical basis to using Keys cubics, but image resampling requires compromise between conflicting requirements, which means that the above "strong theoretical basis" is not a dictate.
« Last Edit: March 28, 2015, 05:00:52 am by NicolasRobidoux »
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 798
    • Hikes -more than strolls- with my dog
Re: A free high quality resampling tool for ImageMagick users
« Reply #428 on: March 28, 2015, 04:42:38 am »

Roughly, RobidouxSharp is the "sharpest reasonable" cubic kernel when used with EWA. (Although some people like the look of EWA Catmull-Rom a.k.a. CatRom, which is even sharper, it is my opinion that this filter gives results which drift too far from "faithfulness". In my book, RobidouxSharp marks the end of "generally usable for EWA".)
EWA Robidoux is a good match for the Nyquist content of gently demosaicked Bayer source.
Even though EWA Robidoux is very popular for downsampling, there is some justification for using something slightly sharper, but not as sharp as RobidouxSharp. EWA Mitchell is a good start.
You could also directly choose the C value using "-define filter:c=value -distort Resize". TTBOMK, you don't need to specify that you want a Cubic filter because the default (Robidoux) is one. In other words, the above should give the same as "-define filter:c=YOUR_CHOSEN_VALUE -filter Cubic -distort Resize". You can also set the "b" value instead of the "c" value, in which case the usable range is 1 (Spline = very soft) to 0 (CatRom = very sharp). "B" is amount of "blurriness" (easy enough to remember).
Terminology: EWA = Elliptical Weighted Averaging = what you get when you use "-distort Resize" instead of "-resize" with ImageMagick. It's also the basis of the "Jinc" methods of madVR. EWA is actually slowly spreading. I think now available for Blender and a few other FLOSS image and video processing programs.)
P.S. Some people like the look of BC-cubics that are not Keys (for which 2B+C=1). In particular, methods with B+C=1 are popular. There is a strong theoretical basis to using Keys cubics, but image resampling requires compromise between conflicting requirements, which means that the above "strong theoretical basis" is not a dictate.

Thanks, Nicolas.  Are MTF curves even a good way to evaluate this stuff?
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #429 on: March 28, 2015, 04:49:56 am »

Unless you really know what you are doing, I am pretty sure that you should not test any scheme that does not filter linear light for downsampling. (Sharpening, with USM, say, is a different story. You can use gamma light for that.)
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #430 on: March 28, 2015, 04:53:50 am »

Thanks, Nicolas.  Are MTF curves even a good way to evaluate this stuff?
They are very informative, but they do not tell the whole story. I read you report with great interest and am still thinking about it.
If there was one single objective function ("quality index") that correctly characterized "image quality", we would take a parameterized filter bank and blindly optimize. There is no such objective function. The door is not closed to "craft".
P.S. In other words, an absurd MTF is probably a strong hint that the method sucks. But there is more to faithfulness than a nice curve.
« Last Edit: March 28, 2015, 05:03:23 am by NicolasRobidoux »
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 798
    • Hikes -more than strolls- with my dog
Re: A free high quality resampling tool for ImageMagick users
« Reply #431 on: March 28, 2015, 05:15:04 am »

P.S. In other words, an absurd MTF is probably a strong hint that the method sucks. But there is more to faithfulness than a nice curve.

Makes sense.  I wonder if Bart's 'downsample' script plays nice with data that is 16-bit linear to start with?  If not I would have to remeasure it and RobidouxSharp.

:downsample
SET sharpAmount="?"
SET /p sharpAmount= Sharpening amount? (none=0, normal=50+, default=100):
IF /i %sharpAmount% == 0 GOTO nodownsharp
IF %sharpAmount% EQU "?" SET sharpAmount=100
ECHO.
ECHO Creating the Down-sampled image with output sharpening, please wait ...
convert ( -quiet "%~1" -type TrueColor -write "%~dp1IMsource.icc" -depth 16 -set colorspace sRGB -colorspace RGB ^
 -define filter:c=0.1601886205085204 -filter Cubic -distort Resize %mSize% ) ^
 ( -clone 0 -gamma 3 -define convolve:scale=^%sharpAmount%%%,100 ^
 -morphology Convolve DoG:3,0,0.4981063336734057 -gamma 0.3333333333333333333 ) ^
 ( -clone 0 -define convolve:scale=^%sharpAmount%%%,100 ^
 -morphology Convolve DoG:3,0,0.4806768770037563 ) ^
 -delete 0 ^
 ( -clone 1 -colorspace gray -auto-level ) ^
 -compose over -composite ^
 -set colorspace RGB -colorspace sRGB ^
 -quality 100 -sampling-factor 4:4:4 ^
 -define png:preserve-iCCP ^
 -profile "%~dp1IMsource.icc" -compress Lossless "%~dpn1_%mSize%%~x1"
« Last Edit: March 28, 2015, 05:22:00 am by Jack Hogan »
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #432 on: March 28, 2015, 05:18:46 am »

Jack: One thing your report makes clear (suggests, really...), is that Photoshop's bicubic is a good, balanced, scheme. (What a surprise.)
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #433 on: March 28, 2015, 05:46:50 am »

Makes sense.  I wonder if Bart's 'downsample' script plays nice with data that is 16-bit linear to start with?  If not I would have to remeasure it and RobidouxSharp.
Not sure. Because of the gamma conversion, you may need a version of ImageMagick that computes in floating point. (Normally, ImageMagick uses 16bit integers as internal storage, which may be a bit thin with a script that does gamma conversion and is fed to frequency analysis.)
It's probably just fine as is. But I'm not sure.
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 798
    • Hikes -more than strolls- with my dog
Re: A free high quality resampling tool for ImageMagick users
« Reply #434 on: March 28, 2015, 05:51:10 am »

Jack: One thing your report makes clear (suggests, really...), is that Photoshop's bicubic is a good, balanced, scheme. (What a surprise.)

Right, on the other hand it looks like the script assumes that incoming data is sRGB and it isn't, so it appears that I need to rework the measurements.  I have changed the last two graphs above to reflect the following code (RobidouxSharp as an example):

convert ( -quiet "%~1" -write "%~dp1IMsource.icc" ^
 -filter RobidouxSharp -distort resize %mSize% ) ^
 -quality 100 -sampling-factor 4:4:4 ^
 -profile "%~dp1IMsource.icc" -compress Lossless "%~dpn1_%mSize%%~x1"

Will this do if the input is 16 bit linear data?

Note that the high frequency 'hovering' has disappeared in the new 'linear' plots above.  It was obviously introduced by the incorrect sRGB <--> RGB round trips.  Here is the original (apparently incorrect) graph for reference, should anyone be interested



Jack
« Last Edit: March 28, 2015, 06:11:54 am by Jack Hogan »
Logged

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #435 on: March 28, 2015, 05:58:52 am »

convert ( -quiet "%~1" -write "%~dp1IMsource.icc" ^
 -filter RobidouxSharp -distort resize %mSize% ) ^
 -quality 100 -sampling-factor 4:4:4 ^
 -profile "%~dp1IMsource.icc" -compress Lossless "%~dpn1_%mSize%%~x1"

Will this do if the input is 16 bit linear data?
It should. TTBOMK, ImageMagick will process the data literally "as is", which means that it will leave (16 bit) linear light as (16 bit) linear light. Note that a lot of the commands are to accommodate jpg, png and a reasonable variety of TIFF. If you know you're using png or TIFF (Thousands of Incompatible File Formats) without destructive compression, you can drop "-quality 100 -sampling-factor 4:4:4 ^" (there, really, for JPEGs or at least JPEG compression) and "-compress Lossless" (probably a good thing to leave it in for TIFFs, but not needed for png).
Warning: All my "ImageMagick syntax comments" are off the top of my head. I've used it a lot. But to be safe I should RTFM. And I'm not.
P.S. Note that the above should pretty much only be used to produce TIFF, because if you output to, say, JPEG or PNG, most image viewers will ignore the ICC and assume it's sRGB (especially true with JPEG, less so with PNG). It's to avoid such complications that Bart's script specifically assumes sRGB as input and output.
« Last Edit: March 28, 2015, 06:25:22 am by NicolasRobidoux »
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: A free high quality resampling tool for ImageMagick users
« Reply #436 on: March 28, 2015, 06:09:59 am »

Thanks for this, Nicolas and Dogway.

Just to throw in a slightly different perspective, I decided to measure MTF curves of a downsized ISO100 D800e slanted edge using the different methods discussed in the latter part of this thread.  The original image is a 300x400 pixel crop directly off DSC_6483.NEF (I downloaded it from dpr and so can you if you so wish) saved as a tiff with absolutely zero processing other than CFA normalization.  I think it's as close as one can get to the actual spatial resolution information captured by the camera.

Hi Jack,

Yes, a slanted edge can reveal lots of useful information which is why I use it a lot, but it doesn't give a complete picture. It e.g. can only suggest that aliasing is possible (or not), but it doesn't know if the original subject had enough contrast to even have a modulation that can cause aliasing. A high contrast edge will survive, but say tiny detail (and small features in digital imaging by definition equal low MTF modulation) may suffer enough loss of contrast due to (re)sampling to become unrecoverable, even with deconvolution. Think about detail with 1% subject contrast (still detectabe by humans), and an MTF response of 10%. That would render the detail at 0.1% modulation, which might be below the quantization/noise threshold.

The MTF view also doesn't reveal the sensitivity for the generation of blocking artifacts or ringing artifacts. Especially the EWA types of resampling (-distort Resize) behave in a very 'organic' way, which avoids some of these types of artifacts much better than tensor type of resampling (-resize). The traditional tensor type of resampling has a higher diagonal resolution, which is nice, but it does cause trouble when we start to push things to the limit. That's why I additionally test with a zoneplate (rings) target, and a star target, and challenging images such as my windmil sample image.

The D800e edge crop does have a bit of a ragged, zipper like, edge structure which is something to be wary about, but I don't think it is too much of an issue for downsampling to 25% or less of the original size. Still, for more objective tests I'd prefer to base my conclusions on a really smooth and sharp edge. One can even compare edges with different contrast, to better judge some (unwanted) effects for really high spatial detail. I can easily generate such edges if needed, at any required angle, but the ISO standard uses an arctan(0.1)=5.71 degrees angle so that's what I usually make (a 1/10 pixel slope is also easy for analysis of CGI versions, camera shots are usually rotated slightly and their slope must be measured).

Quote
That beautiful edge was first measured with MTF Mapper - then downsized 4:1 using Photoshop's standard methods and re-measured.
[...]
The original is soaring at those lofty heights because one of its pixels corresponds to four of the others'.  Nearest neighbour tracks it well, which is good at the lower frequencies, not so good beyond Nyquist because the ideal SFR/MTF curve for a perfect downsizing would track the original's up to Nyquist - but show little or no energy above 0.5 cycles/pixel, the point at which aliasing and moirè start rearing their ugly heads. Bilinear and bicubic show different amounts of attenuation, with bicubic looking like the best compromise.

 The blue curve below shows the upper boundaries of what I think an ideal MTF curve should look like after perfect downsizing only (no extra sharpening), with bicubic for reference as the dotted line.

True, but again watch out for bicubic's (poor) blocking and ringing behavior, therefore also other tests are required.

Quote
Next I plotted the MTF curves from three downsizing algorithms discussed in this thread ('D' option in the script): 'nodownsharp' quadratic, 'downsample' V1.22 and RobidouxSharp.  None were sharpened after downsizing, with 'downsample' set at a minimalist 1% DoG.  Quadratic and downsample 1% fare much worse than our benchmark bicubic while RobidouxSharp does a good job early on and bests it up to Nyquist.  Note, however, how all three have an unwelcome tendency to hover on in the higher frequencies, letting through faux energy there, a fact potentially indicative of the high frequency trouble discussed earlier in the thread.

Yes, but keep in mind that lower contrast small detail will have low MTF to begin with, so down-sampling will only reduce the chance of it having a meaningful modulation after down-sampling. Which is also why I use deconvolution sharpening without too much hesitation. Some of the noise is probably unrecoverable already, while still meaningful detail modulation will be boosted.

Quote
Looking at the MTF curve of RobidouxSharp one would almost feel like throwing a sharp low-pass at its output past Nyquist.  What would that do to the frequencies below Nyquist?  Would it kill its slim advantage over bicubic there?

That's the choice I made after studying the other artifacts. It seems to be better to down-sample a bit soft and deconvolve, than to down-sample sharp and blur/convolve. The sharper down-sampling generates to many compromise artifacts, and resampling generally requires a bit of sharpening to restore some resampling losses but that's not good when there already are artifacts present.

 
Quote
Lastly I decided to give Bart his due and show 'downsample' V1.22 as it was meant to be, with full 100% DoGs.

Just a small addition, 50% is a neutral 1 iteration deconvolution, 100% is default because it adds a bit of extra sharpening that often is visually pleasing.

Quote
The 1% dotted red curve is the heavily attenuated result of its downsizing component.  With 100% 'sharpening' the DoGs component (solid red line) attempts to make up for the loss of energy in the desirable frequency range by amplifying the attenuated curve (and everything else including unwanted noise) back to where it thinks it should be.  I can't help but wish that both the attenuation and the amplification not be so drastic.  Even so the method does not quite achieve bicubic's apparently effortless performance.

The benchmark seems hard to beat when looking at these curves off grayscale raw data.

Thoughts?

Keep the earlier comments in mind. Lower contrast detail may already be lost, so the potentially risky boost of aliasing prone spatial frequencies may be relatively harmless. Also check for other artifacts, the rings target is cruel enough to reveal such poor behavior in 2D. Gamma effects also play a big role in generating or suppressing artifacts.

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

NicolasRobidoux

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 280
Re: A free high quality resampling tool for ImageMagick users
« Reply #437 on: March 28, 2015, 06:13:46 am »

Note that the high frequency 'hovering' has disappeared.  It was obviously introduced by the incorrect sRGB <--> RGB round trips.
It was worse than that: It was converting the linear RGB to linear RGB from the get go as if it was sRGB to start with. One way trip that was never undone, and also got in the way of the mild noise/quantization cancellation one gets for free when convolving in "true" linear light.
P.S. I actually was wondering where the "high noise floor" was coming from.
« Last Edit: March 28, 2015, 06:26:44 am by NicolasRobidoux »
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 798
    • Hikes -more than strolls- with my dog
Re: A free high quality resampling tool for ImageMagick users
« Reply #438 on: March 28, 2015, 06:14:48 am »

Bart, please note that I apparently incorrectly fed your script 16-bit linear data, so I reworked the last two graphs in the original post to reflect the data type.  Results are quite a bit different as you can read above.

Jack
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 798
    • Hikes -more than strolls- with my dog
Re: A free high quality resampling tool for ImageMagick users
« Reply #439 on: March 28, 2015, 06:29:20 am »

Just a small addition, 50% is a neutral 1 iteration deconvolution, 100% is default because it adds a bit of extra sharpening that often is visually pleasing.

Since I have the stuff out, it's easy enough to do one more. Here is 'downsample' with 50% sharpening for reference

« Last Edit: March 28, 2015, 06:34:02 am by Jack Hogan »
Logged
Pages: 1 ... 20 21 [22] 23 24   Go Up