Professional Documents
Culture Documents
http://www.naturescapes.net/phpBB3/viewtopic.php?f=2&t=140615
888-667-0559 / 410-239-8025
8:30AM - 4:30PM Mon.-Fri. EST
Forums
Portfolios
News
Articles
Workshops
Events
Store
About Us
Contact Us
Membership
Affiliate
Links
A post in the Educational Resources forum at Art Morris' Birdphotographers.net mentioned a sharpening algorithm called Richardson-Lucy Image deconvolution. Evidently, this algorithm has advantages over the Photoshop's USM in some situations, and was used to correct for the Hubble telescope's myopia. The algorithm is available from Images Plus. Does anyone at NSN have any experience with this as a part of their work flow? Tom Top
(#p1423062)
by StephenFitzpatrick on Tue Oct 14, 2008 7:31 pm
I don't know what algorithm it uses, but PS's "smart sharpen" is reputed to do deconvolution. Top
(#p1423098)
by ejmartin on Tue Oct 14, 2008 9:07 pm
I've not used RL deconvolution as part of my regular workflow, but I am familiar with the principles involved; and I do use Smart Sharpen, which as Stephen says is reputed to do deconvolution, though I've not seen any definitive statement from Adobe or the people involved. A number of programs can do various forms of deconvolution: Focus Magic uses some form of deconvolution (unspecified). IRIS implements several deconvolution algorithms (but is mainly geared toward astrophotography). Images Plus also implements several algorithms and is geared for astro. Raw Therapee is a freeware raw converter that implements a form of RL deconvolution. Iridient Raw Developer is a not free, Mac only raw converter that implements a form of RL deconvolution. There are almost certainly more. The basic idea behind deconvolution is that image blurring spreads light rays converging on a focus point. The blur can be thought of as convolving a smearing function with the image, in a manner similar to gaussian blur (which is the convolution of a "bell curve" with the image, so that point objects are smeared over the radius of the gaussian). If one knows the precise form of the smearing function, one can mathematically invert the smearing process; in other words one undoes the convolution, which is why it is known as deconvolution. There are several problems in practice: 1. One often doesn't know the precise form of the smearing function, which produces errors in the deconvolution as one tries to make a best guess as to what it was. 2. Noise interferes with the deconvolution process, misdirecting it. Noise gets amplified in much the same way it does with USM and other sharpening routines. 3. The deconvolution procedure can result in "ringing" artifacts similar to the halos of USM, but further from edges and often repeating in a regular pattern. There are algorithms for doing a best fit for the smearing function, and approaches to mitigate the effects of noise and the appearance of ringing; but they are approximations at best and not a cure all for getting an in focus image to begin with. That said, with some understanding of the process and how to tweak it, one can get results that are far better than USM. I don't know of any comparisons between Smart Sharpen and other deconvolution algorithms. It would be nice if Adobe were a little less cagey and provided some
1 of 3
1/13/2012 9:47 PM
http://www.naturescapes.net/phpBB3/viewtopic.php?f=2&t=140615
information about what Smart Sharpen is doing under the hood; that would enable optimization of its use. Top
I have used Focus Magic and I have found it useful once in a very specific set of circumstances. There was a very slight bit of back-focus on an image that I really could not hope to produce again. The image was full frame, so the was still plenty of image data to bring it back from. Although Focus Magic has two functions "correct motion blur" and "correct focus blur", I used to latter. I would say the resulting image is about 90% of what it could have been if the focus was correct. It is clearly different than USM and I doubt anyone here could look at the tiff at full res and tell that something was modified. The "correct motion blur" works fairly well in motion blur situations where the direction of the blur is well documented such as when there is a catchlight that has been streaked in the subject's eye. Of course this is also limited by the resolution of the image and the degree of motion blur. Top
(#p1423960)
by philw on Thu Oct 16, 2008 2:18 pm
Wikipedia's pretty good on some of that stuff. The specific function seems to be concerned with the situation where you know what you're trying to correct for, hence you could maybe use it in something like a bent mirror in the Hubble (although the ultimate fix for that involved hardware I read). I'm not sure how well that would work for focus issues. USM is very crude. The "smart sharpen" is much more sophisticated. Top
I just came across this topic while googling Richardson-Lucy. I have discovered this method thanks to Raw Therapee (and discovered Raw Therapee because it supposedly uses the same deconvolution as Raw Developer). I am extremely impressed with this feature in Raw Therapee, and when used properly it can produce an amount of details I think is unmatched by any other more traditional sharpening techniques. I did try filters like Focus Magic in the past but was never impressed. Smart Sharpen, which is supposed to use some sort of deconvolution -an unblind one I hear- is ok, but falls short when compared to Raw Therapee. I see progress is being made on improving the deconvolution approach (which is great given the papers of Richardson and Lucy were published in the early 70's!). These guys for example seem to have improved both the blinded and unblinded approach: http://www.cse.cuhk.edu.hk/~leojia/proj ... raph08.pdf (http://www.cse.cuhk.edu.hk/~leojia/projects/motion_deblurring/deblur_siggraph08.pdf) Take a look at the images on p.8. I think these results are very promising Top
(#p1643847)
by mhecker on Sun Nov 01, 2009 3:13 pm
I own Images Plus and have tried the RL deconvolution routines. For every image on which RL works a little bit better there's another image that USM works better on. It's a mixed bag. The RL deconvolution algorithm is also very computation intensive and is slow. Bottom line, I no longer use it. I use USM and smart sharpen both as needed. Top
All interesting comments and observations. It looks as if RL deconvolution might be another tool on the belt, but one with limited applications. I plan on experimenting with this during one of those winter days when when it's better to be indoors than out. Top
(#p1643923)
by ChrisRoss on Sun Nov 01, 2009 7:47 pm
RL deconvolution is used a lot on atronomical images, I think the thing with those images is either you know exactly what the abberation is (Hubble for example) or you have stars in the image which you can analyse, being a point source you know exactly what they should "look" like and can use them to guide the deconvolution. Top
For every image on which RL works a little bit better there's another image that USM works better on.
2 of 3
1/13/2012 9:47 PM
http://www.naturescapes.net/phpBB3/viewtopic.php?f=2&t=140615
You are right. But for the images it does work on, I found the results (using Raw Therapee) very pleasing - better then I could achieve with the traditional unsharp mask. In fact, I just finished working on an image that I was sure it would have worked good. Was I ever wrong! Switched back to traditional unsharp masking in no time! Top
(#p1644075)
by ejmartin on Sun Nov 01, 2009 11:06 pm
RL and its ilk are a general method of image restoration. Quality of the results depend on the specific implementation. For an analogy, many of the noise reduction tools are based on multiscale methods (wavelets and their ilk), but not all achieve the same results. Top Display posts from previous: Post a reply 11 posts Page 1 of 1
Return to Digital Topics
Sort by
Jump to:
Powered by phpBB 2000, 2002, 2005, 2007 phpBB Group
Airbrush Makeup Artist Certified Artist, Specializing in Bridal Makeup. Willing to travel. www.gsmakeupartistry.com Photo-shop Tutorials Easy To Follow Tutorials. Download or Watch Online & Master Photo-shop PhotoshopRevealed.com Download Image Converter JPEG TIFF GIF RAW BMP JPG PCX etc. Fast & Easy Image Files Converting avs4you.com/
Buy with
Website Use:
Privacy Policy Terms & Posting Guidelines
Email:
We respect your privacy - your email will not be shared or sold.
2012 NatureScapes.Net - All content on this site is copyrighted material as indicated. Unauthorized use or reproduction is prohibited.
Addre ss: 21009 Gunpowder Road, Manchester, MD 21102 | Phone : toll-free 888-667-0559 or 410-239-8025
3 of 3
1/13/2012 9:47 PM