Sharpening settings using softwares

[tomasis]

Hello I got some weeks to play with Rd1. Epson Photo Raw is very simple but easy to work with. One problem I find that when you load up any ERF file with Photoraw and I get impression that the image looks quite sharp and shows such nice textures. But when I save this as Tiff and jpg then sharpness (crispy textures) is gone with those files. Sometimes tonality is messed up. I often envy M8 users that they don't need process files a lot regarding sharpness due lack of the antialiasing filter. I find that edge enchance +1 and unsharp mask 70% works well for me but I'm not yet sure if I go too far or viceversa. It'd be good way comparing wet print with digital but I still don't have a chance to this yet. So how do you use software to get out max lens sharpness without getting any artifacts? At other words, filter away the antiliasing filter based on D100 sensor.

 

[LCT]

Try Fred Miranda's Nikon CS Pro if it still exists.
The medium setting is almost perfect to me.

 

[iml]

I've never yet used a digital camera which produced images that didn't benefit from some USM. It's just life. Provided you don't overdo it, the results are usually fine.

Ian

 

[mfogiel]

Try Noise Ninja or Neat Image

 

[jlw]

Hello I got some weeks to play with Rd1. Epson Photo Raw is very simple but easy to work with. One problem I find that when you load up any ERF file with Photoraw and I get impression that the image looks quite sharp and shows such nice textures. But when I save this as Tiff and jpg then sharpness (crispy textures) is gone with those files. Sometimes tonality is messed up. I often envy M8 users that they don't need process files a lot regarding sharpness due lack of the antialiasing filter. I find that edge enchance +1 and unsharp mask 70% works well for me but I'm not yet sure if I go too far or viceversa. It'd be good way comparing wet print with digital but I still don't have a chance to this yet. So how do you use software to get out max lens sharpness without getting any artifacts? At other words, filter away the antiliasing filter based on D100 sensor.

I'm not so sure your issue is specific to the R-D 1 or PhotoRaw -- it sounds more like the lack of sharpness is coming into the way you are generating your TIFF or JPEG files.

(You can't -- and don't want to -- "filter away the antialiasing filter"; it has very little effect on image details large enough to be resolved by the sensor. Its purpose is to filter out details too small for the sensor to image, so they don't cause moire patterns and color fringing on finely-textured subjects.)​

Now, where your sharpness is going: If you are saving your images at a smaller pixel size than the camera's original data, some of the fine detail will be lost to interpolation. To make the image smaller, the software has to mathematically average the values of a large number of pixels to make them into a smaller number of pixels. This averaging process removes some of the fine differences between pixel values.

Also, your output device -- whether it be a printer, a monitor, or whatever -- does its own interpolation in order to match the number of image pixels to the number of device pixels, and this also causes loss of apparent sharpness. So for example, if you have a 3000 x 2000-pixel image and you view it in an 800x600-pixel window on your monitor, the computer's video display driver is going to average down 6 million pixels into 480, 000 pixels, and some appearance of sharpness will get lost in the mathematics of this.

What all this should tell you is that, unfortunately, there is no single magic setting that will give you the "best" sharpness for every way you might want to view your image! What you need to do is sharpen the image to give the best results for the way you are going to use it: viewing on a monitor, printing at a small size, printing at a large size, etc.

(This is one of the big advantages of shooting in raw format -- you can save your original raw file as your "negative," and generate from it results that you sharpen exactly the right amount for how you are going to use them.)​

The way I normally handle sharpening is to do it either in Adobe Photoshop or in Adobe Lightroom (which is what I use to manage all my raw files.) If I am making an image to post on RFF, for example, I always first reduce it to be no more than 560 pixels wide (which is a size that looks good on RFF) and then I use Photoshop's "smart sharpen" filter at whatever setting makes it look good on my monitor on that size. The Smart Sharpen filter is pretty smart about sharpening details rather than image noise, so the settings are not ultra-critical -- I just pick what looks good.

For making prints, I apply the sharpening settings in Lightroom to get a print that looks sharp. At first you have to do this by making test prints, because that's the only way to see the final effect. After you get a bit more experience, you will know what settings work best for prints of different sizes.

These techniques apply to any software you might use, not just Photoshop and Lightroom. They also apply to any digital images, not just those from the R-D 1; you have to do this with ANY digital camera to get the sharpest results! The basics are that to get the best results, you have to choose sharpening settings tailored to your output, and you have to judge the results by looking at that output (in other words, you can't judge the best settings for a print by looking at your monitor; you have to look at a print, at least until you have had some practice.)

Sorry if this has made it sound more complicated than you want, but in practice it is pretty easy once you have done it a few times and have learned what settings work best for different kinds of output. (One of the things I like about Lightroom is that you can save combinations of often-used settings and apply them with one click.)

The main thing you need to know is that your R-D 1 can produce results that look very, very sharp, regardless of what type of file you save. You just need to make sure you are following the right procedure for the type of output you want, which would be necessary no matter what type of digital camera you used.

 

[charjohncarter]

Here are some from Scott Kelby's book. I received these from someone on Pnet.

Basic- 125/1/3
Soft Subjects- 150/1/10
Portraits- 75/2/3
Moderate- 225/.5/0
Max- 65/4/3
All purpose- 85/1/4
Web- (200-400)/.3/0

I hope these help, they have me. I'm like Ian all digital images need some.

 

[payasam]

Agree with Ian about not overdoing it. JLW's detailed answer is worth digesting.

 

[tomasis]

dudes, thank you for all inputs here!

LCT, it sounds interesting that the plugin does better than a USM if I have to believe the advert I suppose that Fred Miranda include several custom different sharpenings into the plugin?

Ian, I agree with you Even M8 files require slight sharpening, lol

mfogiel, neatimage looks like as a neat application as the name of program indicates it

jlw, thank you much for the input, wooah. I understand very well what are you trying to mean. If it is not case that AA filter affect sharpness, how come that some sensors like as of M8, DMR, medium format produce quite crispy images even unmodified. I might formulated the sentence "filter out AA" at wrong way when I think about that Rd1 files look tack sharp at its own application so it might not be the cause of AA filter. Maybe you have some explanations about sharpness of M8 files. I agree with you that prints need to be checked separately from the monitor so I assume that sharpening settings are relative as other and you said. I got some tips from your input, thanks again.

john, thanks for the numbers. it is what I needed at most to start from the point. I can see how settings are balanced for different purposes. Lower amount, higher radius viceversa.

payasam, saying "dont overdoing" is very easy but doing is another matter

I played with Photoraw a little today. I got idea to grab a screenshot from original ERF and open this at Photoshop and I'd compare a tiff and screenshoot together so I learned that radius setting is playing big role. That is what I missed for long time!

Except wet prints, I thought about why not compare M8 and Rd1 files at technical basis and see differences how those are processed regarding sharpness. Anybody of owners of both m8 and rd1 can tell about it? I still can't get the lack of AA filter out of my head because I know that most DSLR produces soft images as rd1. Is about preference manufacters put in software processing for liking of customers, huh?

Maybe I was just unlucky that I have seen only sharp M8 files and not the same sharp files from rd1 at some galleries
​

 

[jlw]

If it is not case that AA filter affect sharpness, how come that some sensors like as of M8, DMR, medium format produce quite crispy images even unmodified.

There is no such thing as an "unmodified" digital image! The sensor only outputs a set of numbers that define charge levels. To turn those numbers into an image you can view, the in-camera and/or post-processing software has to do such things as correct for the sensor's non-linear response to light; calculate color values based on the values of nearby pixels; compensate for the sensor's electrical noise characteristics; restore edge contrast lost in the color and noise calculations; etc., etc.

So, even a "raw" image is the product of a lot of calculations, and every manufacturer tries to strike a compromise among the different calculations that will yield the most appealing result to the target consumer. One manufacturer might decide to emphasize low shadow noise at the expense of edge sharpness, another might emphasize edge sharpness and give up some tonal separation, and so on.

At work I use an older medium-format back on a Fuji GX680 camera body. This back is of the "tethered" type -- it contains no image-processing electronics at all, just enough to extract the charge values from the sensor and send them down a cable to a custom interface board in a computer. All the image processing is done by the software that communicates with this interface board. Just to give you an idea of some of the work that has to be done to convert sensor values into images, the software gives you a choice of eight different presets plus user-defined custom values for the amount and type of anti-aliasing alone!

The reason I tell you this is to make clear that it's not just a choice of "more anti-aliasing/less sharpness" or "less anti-aliasing/more sharpness" -- choosing the wrong option for a particular subject can produce LESS sharpness PLUS image artifacts, while the right option can yield an image which is both more crisp-looking and more free of artifacts.

Leica did a good job of selling photographers on the notion that they'd get sharper pictures if the manufacturer would just get rid of that nasty ol' anti-aliasing filter (leaving the faithful to assume that other manufacturers who do use AA filters must be either evil or stupid.) In fact, Leica DOES apply anti-aliasing to DMR and M8 images; they just do it via firmware/software rather than optically, and there is still no way to completely disable it (nor would you want to, because otherwise details beyond the sensor's limit would degrade your results.)

 

[tomasis]

jlw, you know that I have nothing against AA. I sometimes play games so I see how good AA is for producing pleasing images. So you do mean that Leica did the good job to achieve still sharp pictures even using AA through software. Many times I wished that other DSLR manufacters could do same thing excluding hardware AA and using this instead through processing. But for sure manufacters know much more than I how to do things right and sell them for targeted customers. I happen to appealed by images of M8, DMR, newer medium formats so I assume that it is the expense of extensive algorithm processing by manufacters, huh? Is it not that bad exclude the hardware AA filter and using AA instead through software? any disadvantages with this? maybe hardware AA filter is acting as a protection for sensor? I appreciate your knowledge and comments, thanks.

 

[LCT]

...LCT, it sounds interesting that the plugin does better than a USM if I have to believe the advert I suppose that Fred Miranda include several custom different sharpenings into the plugin?...
​

Does better and faster to me but it's a matter of tastes of course.
AFAIC its my only sharpening tool for the R-D1 as well as the Nikon D70.
There are just 5 settings in my version but it's not the latest i guess.
I only use the medium setting (level 2) with R-D1 (and R-D1s) and the low one (level 1) for the D70 due to the weaker AA filter of the latter.
BTW lack of AA filter means moiré and IR problems hence flare, ghost images and oversaturated greens due to the more or less mandatory use of IR cut filters. Not a good deal to me. Would be better to have an accessory AA filter as in some MF backs IMHO.

 

[johnastovall]

I suggest you get a copy of the late Bruce Fraser's book, Real World Image Sharpening with Adobe Photoshop CS2.

It will give you a real understanding of how shaperning works and why a two pass apprach is better than a single pass.

 

[jlw]

So you do mean that Leica did the good job to achieve still sharp pictures even using AA through software. Many times I wished that other DSLR manufacters could do same thing excluding hardware AA and using this instead through processing.

I think that the situation was more that Leica was unable to use an optical AA filter in the M8 because it has a thin body, so the "chief ray angle" (roughly, the angle at which a ray from the center of the lens strikes the edge of the sensor) is very steep. Because the angle is so steep, rays reaching the edge of the sensor would have to pass through much more glass than those reaching the center, and this would cause severe vignetting. It is exactly for this same reason that they were not able to include a sufficiently effective IR filter, and instead have to rely on the user to attach IR-cut filters over the lens.

(Most DSLRs have much thicker bodies, so the chief ray angle is not as extreme and causes no problems in the use of an optical antialias filter and IR-blocking filter. I don't know what the situation is with the DMR, but possibly the Leica R body is thinner than most...)

Anyway, being unable to include these optical elements in the M8 because of its thin body depth, Leica cleverly made a marketing virtue out of this necessity by saying that eliminating the AA filter would make the pictures sharper, and that eliminating the IR-blocking filter would facilitate the use of creative IR effects!

The point I was trying to make in my earlier discussion was not that optical AA is better than software AA or vice-versa, but merely that it is an oversimplification to assume that if camera A's images look sharper than camera B's, it is because of the presence or absence of AA filtering.

Certainly you can decide that you prefer the "look" of the images from camera A or camera B... but you can't reliably infer why, since there is NO way to examine an image file that has not been subjected to some sort of in-camera image processing, making an "apples-to-apples" comparison impossible.

Incidentally, while I have no way of determining whether this is true or not, the makers of optical AA filters claim that they do NOT have any reducing effect on the sharpness of details below the filter's frequency-cutoff limit. Apparently these filters are not simple "softeners," but rely on birefringent crystals that can pass coarser details unhindered while blocking finer details at a specific cutoff point.

AA filter manufacturers also say that the moire and image-artifact effects that are suppressed by an optical filter can not be fully corrected by software alone, because by the time the software "gets" the image, the damage is already done.

As I said, I do not know whether or not these claims are true, but you can read them and decide for yourself. Here are links to two brief technical papers from Sunex, a maker of sensors and AA filters, explaining the points about how an AA filter works and the effects of chief ray angle:

Link to the first paper: Click here.

Link to the second paper: Click here.

 

[tomasis]

jlw, thank you for rescueing me from the cruel Leica (maybe also PhaseOne, Kodak) PR machines The thing with AA sounds more complicated than I thought. I guess that I will dig more into AA issues with MF backs and reasons by manufacters for not using optical AA filter. That was interesting what Sunex said about the perfomance of its own filter.

john, wow the whole book just about sharpening? 30$ is cheap comparing with the whole price of M8.

LCT, I guess that I'll shed some $$ for a silly plugin No, I'm just curious what this can do comparing with USM. It'll be fun

 

[--]

Edge Enhancement translation/conversion factor

Edge Enhancement translation/conversion factor

Hello JLW - I experience the same thing as tomasis - and have read your very thorough explanations a few times. I am still somewhat puzzled. I must haste to add that I have shot digital for four years but the R-D1s is the first camera where I am seriously considering shooting digital instead of film when it really matters.

I do not have a quality printer so this it only based on a comparison between two images (as my eyes/brain interpret them) on the same TFT screen. Just as tomasis I fiddle a little in Epson PhotoRaw and the results before conversion are bang on concerning sharpness (everything really). I would literally like to mount the screen instead of getting a print – to me they are that good. As far as I am capable of following your technical explanation (my mental faculties are not up to snuff at the moment, sorry about that) then when converting the .ERF file to a 16-bit max tiff (13.54m) image the quality should actually be better in the tiff file than in the original .ERF file (or am I mixing camera output file size with software output file size?) since I am doing the opposite of what you are writing here:

Now, where your sharpness is going: If you are saving your images at a smaller pixel size than the camera's original data, some of the fine detail will be lost to interpolation. To make the image smaller, the software has to mathematically average the values of a large number of pixels to make them into a smaller number of pixels. This averaging process removes some of the fine differences between pixel values.

Yet the quality drop (in sharpness only) is immense.

Also, your output device -- whether it be a printer, a monitor, or whatever -- does its own interpolation in order to match the number of image pixels to the number of device pixels, and this also causes loss of apparent sharpness. So for example, if you have a 3000 x 2000-pixel image and you view it in an 800x600-pixel window on your monitor, the computer's video display driver is going to average down 6 million pixels into 480, 000 pixels, and some appearance of sharpness will get lost in the mathematics of this.

Then I tried to make twelwe different conversions as follows:

a)4 jpegs – 2by2 matrix – size (6 resp 13.54 on one axis) and sharpening (none, 5) on the other
b)4 tiffs 8-bit – same blend
c)4 tiffs 16-bit – ditto

When viewed in Windows (sorry) Picture Viewer there is no difference whatsoever based on jpeg or file size – only factor of any visual difference is sharpening (5 being more sharp than 0 but (forgive me the unprecise description) but not very, very sharp)

I would easily believe that the Windows Picture Viewer simply is not up to the job software-wise compared to Epson PhotoRaw. Yet the story is the same in Photoshop Elements – only visual difference is whether the picture has been sharpened or not.

My preliminary conclusion - probably known to all you rest already – is that the internal interpretation in Epson PhotoRaw of sharpening (the rather rough -5 to 5 scale) is far wider than the resulting span when converted into jpeg/tiff. The .ERF file shows a huge difference between 0 and 5 in EdgeEnhancement , whereas the converted jpeg/tiff picture shows far less. It is as if Epson “forgot” to put that multiplying factor X on that would have created the same span in jpeg/tiffs that are in an .ERF with EdgeEnhancement.

I do not know if this makes sense to anybody but me . What I really need is the button to push that employs this factor when converting the adjusted .ERF file into jpeg/tiff which would supply me with perfect, finished pictures (at least for on-screen viewing – printing will be in another lifetime…). But such a feature is not there – is it?

I am keenly aware of the many alternative solutions (including sharpening in Elements) but it just seems such a waste when the PhotoRaw result is so good.

Any comments much appreciated

Regards
Karspoul

PS: I love this camera!:dance: