Important Sharpening Information!

Peter Kes said:

Too bad your picture is not visible (yet).

Click to expand...

Peter,
If you click on the image, does it not show? If it does, try middle click (windows) to show it in a new window.
Roger

opening it in a new window doesn't work either.

Chris

This is all VERY interesting. I have a suggestion that I think most of us would appreciate. How about a REAL side by side test or two starting with an original Raw files? In looking at Rogers comments and his website, it looks like he is brining a bit of "rocket science" to the game.

What if we started with 2 or 3 raw images sent to Roger, Artie, and any other of you PS gurus, and then compared the final results side by side? In addition to posting the results here, we may want to have a link to another website where we can view the full sized end product.

What do you think?

Bill

This topic is very interesting and enlightening. Thanks for taking the time to share all this information.

Hope we can see the images soon.

Diana

I've tried Firefox and IE and none of the right-click options in either show mclark's image for me.

Grace, as the original poster, and owner of the image, with your permission, I'll post the Richardson-Lucy computed image to a temporary location on my web site, then remove it once the image shows up here, so everyone can see it. I guess all the moderators who could approve the image are out for the holiday. So let me know if that is OK.

Roger

I'm weighing in late, here, but I'd like to respectfully disagree with Arthur and add my views...

You can achieve a very much finer, more detailed result by sharpening first, THEN downsizing. You can add a touch more sharpening to the downsized (web) version afterward to crisp it up and overcome both the softening effects of downsampling and also pre-compensate for blurring on some monitors (but watch out for that latter; lots of folks have nice sharp LCDs nowadays).

ESPECIALLY with nature and animals, it's all about details and the perception of details.

And all sharpening methods are not created equal.

Deconvolution (e.g., Richardson-Lucy) can be a very good means to sharpen certain images, and others simply fall apart with it. It's worth noting that astrophotographers use all kinds of deconvolution to overcome the limitations of seeing and optics.

I'm not saying it's impossible to work with images at the size posted on the web. Here, for example, is what I could make of the posted image above. The process I used involved upsampling, sharpening, then re-downsampling back to original size. Compare this to Arthur's version. You simply cannot achieve sharpening this fine without working at a higher resolution. Grace, I hope you don't mind my reposting your image; if you do please let me know and I'll take it down.



As the author of what I'd like to think are best-in-show sharpening tools myself I'd like to see the full-sized image as well.

-Noel

Edit: I just read through Roger's info. He's quite clearly (sorry for the pun) an expert in image processing.

Noel, beautiful work. What did you use for the sharpening?

This has been a very informative thread.

Sid

Noel,
How did you get your image to load?
Roger

It's not loaded on this server; use IMG and /IMG tags with square brackets around them, and paste your URL between.

I used my own fractal sharpening actions for Photoshop, which employ the Genuine Fractals plugin.

-Noel

Noel, are you using Genuine Fractals to do the sharpening when you upsize? If yes, what values do you use. I use GF to scale my photos, but had sharpening turned off.

Thanks.

No, I'm not using the sharpening function inside GF. I use the basic Genuine Fractals function to upsample, then I manipulate and sharpen the image at the upsampled resolution, then finally return the image to original size.

-Noel

All,
My attachment has been approved, so you can see my sharpening on the image. See my post from August 30.
Roger

Looks great, Roger!

-Noel

Hey, to all you who asked permission to work on my image, I say a big thank you for going ahead and doing so! By all means. This was a great thread, I must take some time to digest it, but all of the re-worked images look fantastic.

Thank you Roger and Noel for this enlightenment. You have how-ever, created some new questions that I hope one of you could answer.

Roger, would the Richardson - Lucy method in any way be analogous to PS "Smart Sharpen" / Lens Blur or Motion Blur functions? I have learned to ask you how much things cost, assuming one could purchase it. Can you offer a source for this program?

I have long been an advocate of original image sharpening mostly for workflow purposes also while of course, still maintaing my original TIFF file for any possible opps.

Noel, My California based custom lab (very occasionally used for scans) always poo pooed on GF unless absolutly necessary, but what I am seeing is telling me otherwise and confirming my suspicions. How does GF intergrate with CS3?

Bill McCrystyn said:

Thank you Roger and Noel for this enlightenment. You have how-ever, created some new questions that I hope one of you could answer.

Roger, would the Richardson - Lucy method in any way be analogous to PS "Smart Sharpen" / Lens Blur or Motion Blur functions? I have learned to ask you how much things cost, assuming one could purchase it. Can you offer a source for this program?

Click to expand...

While some people do like smart sharpen, I've never been able to get as good of results as that I get with Richardson-Lucy, Smart sharpen is still a single step process (so is fast) and the math of the blur says there is no single step solution, so I would expect Richardson-Lucy deconvolution should do better. Of course,any of these tools with the wrong parameters can produce bad results. I find with each of them different parts of an image needs different parameters, so I'll try several and bring them in a layers and blend the best areas together.

I'm using ImagesPlus from http://www.mlunsold.com (currently $210). The software is designed for astronomical processing. If you do stacking of night photos it works great too. The interface is not as good as photoshop in my opinion. But it does work in 32 bit floating point and you can push limits much more than in photoshop because photoshop does 15-bit integer math with additive approximations for multiplies, which compromises accuracy.

Thanks for the email alert. I just got back from Alaska and have been processing my bird and bear photos.

Roger

Hi Bill,

Genuine Fractals 5.0 is a plugin for photoshop that allows you to upsample an image via a process where it applies fractal algorithms to match the original image, then extrapolates those algorithms in the upsampling process. The detail that's "created" closely matches that which is found in everyday things. It simply allows you to specify the new size you want for the image and the image size is changed. It can be pretty slow with large images at high bit depth.

I have no business affiliation with OnOne Software (the makers of Genuine Fractals), by the way.

My dSLR Fractal Sharpen actions upsample the image to 250% of original size with Genuine Fractals, which preserves the crispness of subject edges with minimal "ringing" or "haloing", manipulates the image at the higher resolution to add additional edge crispness, and finally downsizes back to original size. This is incredibly effective for that last little bit of sharpening, such as what is needed with most digital cameras. Here are a couple of before / after images to give you an idea of the results... The first is done with one of my most aggressive actions, Heavy Sharpen Low ISO. The second with a less aggressive one, Medium Sharpen Max Texture. I describe the "feel" of the results as "refined detail", and it's well suited for prints.





Like Bill, I am also very familiar with deconvolution sharpening (I process a lot of astrophotos) and there are limits to what the deconvolution can produce. It's highly dependent on data quality for one thing, and personally I find the results from deconvolution can all too easily get to a point where the image looks overprocessed, especially at the edges of subject material. I worked to avoid that look with my own sharpening actions.

Lastly, I have been experimenting with a technique I call "iterative smart sharpening", in which multiple passes of very light Photoshop Smart Sharpen are applied at slightly decreasing radii. This is proving surprisingly effective.

-Noel

Wow... what a great thread. If I thought I had a clue about sharpening this thread has dispelled that myth for me.

Has anyone here used Lewis Kemper's Multi Sharpening method and if so how does it compare to deconvolution, fractal sharpening, and iterative smart sharpen?

Thanks again, both you guys. It would seem you are indeed the Guru's of the sharpening universe. I will be looking on line for a new software plug-in, one that I always suspected would work (back in my old 8mp 20D days). It is interesting Noel that I have been playing around with a similar tactic of decreasing radius passes with unsharp mask. I guess I wasn't entirely barking up the wrong tree. I'm not sure why multiple passes seems to work so well - but it does. Would it be partly because you are averaging down % wise on each pass rather than taking one big chuck off the original. I guess you could look at like moving down in sandpaper grades to get to the final finish.