Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: andyptak on January 18, 2008, 11:19:52 am
-
One of the Stock agenices I use requires a minimum of 48 megs for an 8 bit image. I shoot with Pentax K10D's and a 16 bit TIFF file is roughly 50-60 megs, half that when I drop it to 8 bits. At their suggestion, I'm using Genuine Fractals to bump up the image size and I'm getting a lot of QC rejects for "soft definition". I know the orginals are fine because I have no QC issues with the same shots from other agencies who do not require interpolation.
I'm using the Express option, because I thought it would be simpler and less of a learning curve, but I guess that's been a mistake. Under the option of "Pixel dimension" which works in terms of %, I increase it by whatever amount it takes to get to the file size I need.
Anybody have any experience with this plug in, in terms of settings etc., if I use the full, rather than Express option? Thanks in advance.
-
You're expecting software to fix a problem that can only be solved by capturing more real image data instead of "faking" image information by making educated guesses with software. Some agencies are more stringent regarding the technical quality of their images than others. As you've discovered, trying to fool the more demanding by using software interpolation instead getting a higher-resolution camera has mixed results at best, and futzing with the settings is not going to change that. Capture resolution does make a real difference, or we could all use cell phone cameras and upsize in software to whatever size we needed later.
-
To the contrary. The agency is recommending it's contributors use Genuine Fractals because it takes a $25,000 Hasselblad to naturally produce 8 bit file of a 50 meg minimum. My problem is not getting caught trying to fake it, it's the learning curve of following their recommendations. I gave enough camera information that you should have known I wasn't using a piece of crap. Apology accepted.
-
To the contrary. The agency is recommending it's contributors use Genuine Fractals because it takes a $25,000 Hasselblad to naturally produce 8 bit file of a 50 meg minimum.
Bull$#@*. Both the Canon 1Ds-II and 1Ds-III meet or exceed the minimum file size (16.67MP) with no upsizing. The Mark II can be had for less than $4000 now that the Mark III is out.
All upsizing software like Genuine Fractals can do is spread out captured detail among a greater number of pixels without introducing obvious pixelization artifacts such as jagged edges. You can tweak the settings to make edges crispier, but that will run you afoul of a whole new set if image quality issues (the sharpening artifacts and probably noise, too). If using such programs gets your images rejected for softness, then you are out of luck.
If you're submitting images to an agency with a 50 MB file size minimum, you probably are competing against medium format digital shooters who can reduce their file size by half and still meet the minimum. If you expect to compete under such circumstances, then you should probably switch to a camera that doesn't require you to upsize to meet the minimum requirements. Your camera is not a piece of crap, but you do need to realize that there are DSLR's with twice your pixel count and better per-pixel image quality, and the medium-format shooters have nearly 4 times your pixel count and somewhat better per-pixel quality than DSLRs. Upsizing your images with Genuine Fractals or any other program is not a realistic means to counter the disadvantage you're working under.
Apology accepted.
Your naivete does not constitute grounds for me to issue an apology. Request denied.
-
You're expecting software to fix a problem that can only be solved by capturing more real image data instead of "faking" image information by making educated guesses with software.
Exactly.
That an agency would allow this but disallow a smaller capture device shows how clueless they've become with respect to photography and imaging.
-
If you are using a PC, or an Intel Mac, you might try a deconvolution program called Focus Magic. I don't have the link handy but a Google search will turn up their website. They have a downloadable trial version.
The program is very helpful in countering the blur caused by A-A filters, and by measuring the width of the blur, it can actually reclaim the data within the blur, giving the impression of refocusing the image. The effect is subtle, and very data-intensive but I include it in my workflow for almost every image.
You might have good results if it were used before uprezzing, or possibly after doing so, or, even before and after. It might give your files the crisper look. You have nothing to lose in trying it.
-
One of the Stock agenices I use requires a minimum of 48 megs for an 8 bit image. I shoot with Pentax K10D's and a 16 bit TIFF file is roughly 50-60 megs, half that when I drop it to 8 bits. At their suggestion, I'm using Genuine Fractals to bump up the image size and I'm getting a lot of QC rejects for "soft definition". I know the originals are fine because I have no QC issues with the same shots from other agencies who do not require interpolation.
I'm using the Express option, because I thought it would be simpler and less of a learning curve, but I guess that's been a mistake. Under the option of "Pixel dimension" which works in terms of %, I increase it by whatever amount it takes to get to the file size I need.
Anybody have any experience with this plug in, in terms of settings etc., if I use the full, rather than Express option? Thanks in advance.
That sounds like Alamy, and their recommendation of Genuine Fractals is years out of date. Most people using less than 16MP cameras simply upsize in Photoshop/Lightroom using Bicubic Smoother. If you're using ACR you can do it during conversion by choosing 5120 x 3413 as the output size.
We're still selling pics on Alamy upsized from a 6MP *istD, although they might not get through QC any more. We've only had one K10D pic through QC so far, but there are some more in the queue.
Colin
-
their recommendation of Genuine Fractals is years out of date. Most people using less than 16MP cameras simply upsize in Photoshop/Lightroom using Bicubic Smoother. If you're using ACR you can do it during conversion by choosing 5120 x 3413 as the output size.
The latest version of GF is much improved over previous versions and by my tests slightly outperforms ACR, CS3 or Lightroom for uprezing.
-
The latest version of GF is much improved over previous versions and by my tests slightly outperforms ACR, CS3 or Lightroom for uprezing.
[a href=\"index.php?act=findpost&pid=168909\"][{POST_SNAPBACK}][/a]
That's interesting. Is that at the 1x to 2x order of linear increase? GF used to be better for very large blow-ups, but showed no consistent advantage at these lower orders.
Colin
-
@Travelling Light - Thanks. Nice to hear from someone who's not in a bad mood.
You're right, it is Alamy. Why they do this I don't know, but they have their reasons, I guess. As you've probably guessed, I'm fairly new to digital - been an old film hanger-on for years. My mind is still stuck on, have a decent camera, good lens and a low ISO film, and the rest will take care of itself. Pixels and image dimensions and all of that stuff didn't enter into it - enlargements were all up to the lab and you could blow something up until the film grain ruined it.
Anyway, it's a bit of a learning curve, but I'm enjoying it. Thanks again.
-
That's interesting. Is that at the 1x to 2x order of linear increase? GF used to be better for very large blow-ups, but showed no consistent advantage at these lower orders.
Colin
[a href=\"index.php?act=findpost&pid=168920\"][{POST_SNAPBACK}][/a]
I only tested it for 2x. As I said the difference is slight. It slows down the workflow over doing it in ACR, but for what I do digitally (architecture, primarily for magazines with a 5D), all aspects of the workflow are critical for best results.
-
I've got some screenshots showing the extreme of onOne's GF effects on a 4mp image here (http://www.plugsandpixels.com/genuinefractals.html). I haven't done print tests or intentional head-to heads yet, but the same sort of application can be seen for Alien Skin's BlowUp here (http://www.plugsandpixels.com/blowup.html).
Other related resampling options, some with similar screenshots, are listed here (http://www.plugsandpixels.com/imageenhancement.html) under "Resampling/interpolation".
-
I've got some screenshots showing the extreme of onOne's GF effects on a 4mp image here (http://www.plugsandpixels.com/genuinefractals.html). I haven't done print tests or intentional head-to heads yet, but the same sort of application can be seen for Alien Skin's BlowUp here (http://www.plugsandpixels.com/blowup.html).
Other related resampling options, some with similar screenshots, are listed here (http://www.plugsandpixels.com/imageenhancement.html) under "Resampling/interpolation".
[a href=\"index.php?act=findpost&pid=169343\"][{POST_SNAPBACK}][/a]
Screen shots are basically useless. Print it (and test it next to a Photoshop BiCubic smoother upsize).
-
The latest version of GF is much improved over previous versions and by my tests slightly outperforms ACR, CS3 or Lightroom for uprezing.
[a href=\"index.php?act=findpost&pid=168909\"][{POST_SNAPBACK}][/a]
Hi,
I concur with your findings when comparing GF to PS CS3 uprez.
It's subtle but there is a nice quality about the GF that side by side makes it the winner.
So long for now, TOM
-
Screen shots are basically useless. Print it (and test it next to a Photoshop BiCubic smoother upsize).
[a href=\"index.php?act=findpost&pid=169346\"][{POST_SNAPBACK}][/a]
Andrew,
Could you explain why?
If I am viewing each uprez at 100% and letting PS CS3 Match Zoom and location?
Thanks, and so long for now, TOM
-
Andrew,
Could you explain why?
If I am viewing each uprez at 100% and letting PS CS3 Match Zoom and location?
Thanks, and so long for now, TOM
[a href=\"index.php?act=findpost&pid=169380\"][{POST_SNAPBACK}][/a]
Cause like rubber hitting the road, you have to print out the differences (there's no reason to upsize files to this size just to view them).
In the examples I see of the different interpolation methods above, I see one side that shows very well defined but "pixelated" rendering and one that looks smooth and smeared. The later I suspect is using the GF upsizing methods. Neither looks good. They both look different. How do they look when we do what we'd eventually do with the upsized files, print them out?
-
What you see at 100% on a monitor is not exactly how the print will look. For a good print, one usually needs to sharpen to the point that the file looks somewhat "crunchy" or oversharpened onscreen. The same applies to upsizing software. Artifacts that look unpleasant on the monitor at 100% may look fine when printed, or vice versa. The only way to tell what really works best is side-by-side print comparisons.
-
Cause like rubber hitting the road, you have to print out the differences (there's no reason to upsize files to this size just to view them).
Hi Andew,
OK, but can you consider the following? (Please -- I know this may be less scientific and more subjective and narrative, but bear with me). Thanks.
My workflow is to take my file and uprez with each method (with PS CS3 (bicubic or bicubic sharper), Blow up, and GF) and then use PK Sharpener as I usually would on the uprezzed files. I then examined them alongside each other (using PS CS3 Match & Zoom) at 100%, and especially examine any area of importance vital to the success image.
I did this work in the same "spirit" as soft proofing, knowing that the soft proof is not perfect, but looking for the "pop." With soft proofing it's usually subtle color effects. With this uprez/sharpening comparison, I am looking for the image that has more, for lack of a better word "life."
Because we are dealing with rather large images 36x30 inches or larger, and to print out all these versions to arrive at a conclusion would be very costly in time consuming and materials, we thought we had a good solution to this problem.
Are you saying, I really have a faulty workflow, and have made poor choices that really aren't valid because until the "rubber hits the road" you cannot really know what's happening by looking at the screen at 100%?
(ASIDE: Over the past 9 years, I have witnessed that the "uprez" solutions evolve and change. There have been times at which different solutions work better than others, but I have been comparing these on the screen at 100%, and thought that was a safe way to go. I fully understand that at different magnifications Photoshop is aliasing the files in ways that would negate such an approach. But I thought 100% was the real deal. Am I wrong about that?
Thanks, and so long for now, TOM
-
Are you saying, I really have a faulty workflow, and have made poor choices that really aren't valid because until the "rubber hits the road" you cannot really know what's happening by looking at the screen at 100%?
Well yes, unless your final output is a HUGE upsized image you want to view on a display (doubtful).
You can print out an 8x10 section of each, no need to output both to an enormous print.
And in the end, there's no substitute for real data. These products are crutches or to be used when there's a gun to your head.
For fun, take the original capture resolution and output an 8x10 section. Then downsize that a great deal, and upsize back using the products. I think you'll see the don't come close to actual captured data.
-
Andrew's got good points, of course. I think of interpolation software as "hamburger helper".
As for showing their capability on a website (such as I do), I chose to use extreme examples for visibility's sake. I wonder if the more realistic printed out side-by-sides would be as visible once scanned and posted online? I don't tend to make prints (I mainly do web and PDF publishing).
-
My workflow is to take my file and uprez with each method (with PS CS3 (bicubic or bicubic sharper), Blow up, and GF) and then use PK Sharpener as I usually would on the uprezzed files.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=169524\")
Another approach is the use of a deconvolution algorithm such as the Adaptive Richardson-Lucy Iteration rather than or in addition to the unsharp filter. Whereas sharpening merely creates the illusion of sharpness, the R-L algorithm actually restores sharpness. See [a href=\"http://www.clarkvision.com/imagedetail/image-restoration1/index.html]Roger Clark's[/url] web site. Roger reports that he can get 2x (linear, 4x area) larger prints with this method. Disadvantages of the R-L are increased compute time and increased noise.
The smart sharpen filter of Photoshop is another deconvolution algorithm if you can figure out how to use it.
Bill
-
You can print out an 8x10 section of each, no need to output both to an enormous print.
And in the end, there's no substitute for real data. These products are crutches or to be used when there's a gun to your head.
[a href=\"index.php?act=findpost&pid=169527\"][{POST_SNAPBACK}][/a]
Hi Andrew,
Good points. In fact I usually do print out test strip like examples that I compare side by side, if the project warrants that.
Massaging and squeezing pixels to give up their inner selves is our game. I guess I should have mentioned that a lot of the images I might be using the "hamburger helper" on are in fact synthetic images to start. We create machine art with fractal programs, and synthetic art with Studio Artist, and for more traditional looks it's the wild and wooly Painter X with millions of pixel dripping choices at hand, and of course Photoshop and all the tools. Sometimes the image began as any and all of the above and not always at the right size or resolution. I should have made clear I am not simply uprezzing digital photographs or worse yet jpgs with the intent of faking great, crisp highly detailed photographs. Digital art is an interesting area to work in and the tools mentioned sometimes make effects that are successful.
As far a photography goes, I have shot traditional film from half frame Tri-X all the way up to 8x10 Fujichromes, but that is like a ghost in the machine now. I have also done Photoshop retouching (10-12 long years ago) when the files were provided in CMYK and were already sharpened by? at? the scanner. We have come a long way! Retouching those sharpened CMYK drum scans was quite interesting, and so backwards. After reading Real World Camera Raw, Image Sharpening by Bruce Fraser, and your book Color Management for Photographers, I must say I owe a lot to you all (would that be the Pixel Mafia?) and in the end, it's a vision thing.
So long for now, TOM
-
Pixel Mafia???
hehe . . . .
-
A synthetic image is going to be even more an apples to orange test but yes, print em out.
-
... The smart sharpen filter of Photoshop is another deconvolution algorithm if you can figure out how to use it.
Here's Forensic Photoshop's summary take on it:
"Monday, October 15, 2007
Smart Sharpening - Fixing focus issues the easy way
In the Forensic Photoshop work flow, the first step after acquiring the image is to address issues of focus. One of the easiest ways to deal with focus is with Adobe's deconvolution filter, also known as Smart Sharpen. Here is a quick overview of the settings:
"Filter>Sharpen>Smart Sharpen
"There are two ways of working with the Smart Sharpen interface, Basic and Advanced. The Advanced mode has the same settings as Basic and allows you to fade the effect in the shadows and/or highlights.
"Amount: Like Unsharp Mask, a higher value increases the contrast between edge pixels, giving the appearance of greater sharpness.
Radius: How far out do you want to apply the effect? The greater the radius value, the wider the area that will be subject to an increase in contrast. Enter too large a number and you will have noticeable halos.
Remove: Offers a selection of three ways to find and minimize the blur, thus sharpening the image:
Gaussian Blur - This option works like the Unsharp Mask filter and is the fastest.
Lens Blur - This option does a great job looking for edges and detail and thus produces finer sharpening of detail.
Motion Blur - Got camera or subject motion blur? This option looks for it and reduces the effect. Be sure to measure the angle of blur with the Measure Tool, if you plan to use the Motion Blur option. (Make note of the measured value in the Options bar before entering the Smart Sharpen interface, and then input it into the Angle control field).
More Accurate: Processes the file twice for better removal of blur and enhanced sharpness. Be warned, this option can double processing time.
"From the Advanced Settings:
Fade Amount: This adjusts the amount of sharpening in the highlights and/or shadows. Fade it all the way (100%) to conceal the sharpening. No fade (0%) gives maximum sharpening. Think of it as a Blending Mode's opacity slider in reverse.
Tonal Width: This controls the range of tones that are modified by Smart Sharpening. The lower the setting the more subtle the result.
Radius: The slider works like Radius in Basic mode and controls the size of the area around each pixel that is used to determine whether a pixel is in the shadows or highlights.
"Now that you've seen how each of the settings works, give it a try on your images. Input your images. Then correct the focus. Then look at contrast, lighting, and colour. Then deal with noise. Then sharpen edges and detail. Interpolate if necessary and add a final touch of sharpening before you print or archive.
"There are some photographers who advise moving into LAB mode before using Smart Sharpen. I tend to agree, but with certain exceptions and it depends on the image. Mostly, I would convert to LAB mode during the final sharpening step, not necessarily at the focus correction step. Try it for yourself and see if you can tell the difference."
— <http://forensicphotoshop.blogspot.com/2007/10/smart-sharpening-fixing-focus-issues.html>
... about which they also had to say elsewhere:
"Almost all images are soft upon arrival into Photoshop. In class, we used the Smart Sharpen filter to address our focus problems. We saw that choosing Gaussian Blur as our removal method turned the filter in a fancy Unsharp Mask filter (remember the modern art masterpiece we created as a result of complimentary colour fringing at the edges). We also noted that in order to use Motion Blur as our removal method, we had to first determine the angle of the motion and put this value into the angle box. With these things in mind, we chose to use Lens Blur as our removal method, fixing focus with Smart Sharpen, Photoshop's built-in deconvolution filter. We also noted that there were some plug-ins that could help us out; namely Optipix <http://www.reindeergraphics.com/index.php?option=com_content&task=view&id=28&Itemid=51> and Clear ID <http://www.oceansystems.com/dtective/clearid/index.html> (both written by Chris Russ <http://screen-online.de/blog/archives/000016.php>)."
— <http://forensicphotoshop.blogspot.com/2007/11/leva-pre-conference-training-post-game.html>
-
Here's Forensic Photoshop's summary take on it:
"Monday, October 15, 2007
Smart Sharpening - Fixing focus issues the easy way
In the Forensic Photoshop work flow, the first step after acquiring the image is to address issues of focus. One of the easiest ways to deal with focus is with Adobe's deconvolution filter, also known as Smart Sharpen. Here is a quick overview of the settings:
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=169761\")
George,
Thanks the information on the general purpose of the smart sharpen controls. The problem is to choose the right settings among hundreds of combinations in a rational way rather then by trial and error with real time on screen feedback. As you know, the trick of using deconvolution methods is to determine the point spread function that is used in the deconvolution to undo the blurring. In the case of motion blur, this can be relatively simple. In the case of defocus, one method recommended by [a href=\"http://www.focusmagic.com/]Focus Magic [/url] is to examine a point specular highlight and determine the diameter of the blur circle in pixels and use this for the deconvolution. I'm not sure how you would do this with Smart Sharpen.
Of course, the image may have more problems than defocus, such as the image softening introduced by the camera's blur filter. Bruce Fraser's work flow uses the unsharp filter with a radius determined by the camera resolution and amount determined by the strength of the blur filter to restore the appearance of sharpness (sharpening for source). Jonathan Wienke and others have reported very good results for source sharpening with deconvolution (Jonathan uses Smart Sharpen).
Bruce Fraser wrote a whole book on sharpening, revolving mostly around the unsharp mask. Its optimal use is pretty complicated. He discussed the Smart Sharpen filter, but was not enthusiastic about it--perhaps he hadn't yet figured out how to use it to the greatest advantage. What would be most helpful would be a source describing these newer methods and similar to Bruce's book.
Bill
-
The problem is to choose the right settings among hundreds of combinations in a rational way rather then by trial and error with real time on screen feedback. As you know, the trick of using deconvolution methods is to determine the point spread function that is used in the deconvolution to undo the blurring. In the case of motion blur, this can be relatively simple. In the case of defocus, one method recommended by Focus Magic (http://www.focusmagic.com/) is to examine a point specular highlight and determine the diameter of the blur circle in pixels and use this for the deconvolution. I'm not sure how you would do this with Smart Sharpen.
Of course, the image may have more problems than defocus, such as the image softening introduced by the camera's blur filter. Bruce Fraser's work flow uses the unsharp filter with a radius determined by the camera resolution and amount determined by the strength of the blur filter to restore the appearance of sharpness (sharpening for source). Jonathan Wienke and others have reported very good results for source sharpening with deconvolution (Jonathan uses Smart Sharpen).
[a href=\"index.php?act=findpost&pid=169943\"][{POST_SNAPBACK}][/a]
The controls in Focus Magic are simple, and with a little practice it becomes easy to choose the optimum blur width to use for an image. The before/after preview windows in FM show the result (for OOF blur, which includes softening by the AA filter) at 100%. You can click on the least blurred part of the image and FM will suggest what the blur width is, in pixels. You can also reduce or increase the application of the effect to more than or less than 100% (almost never needed, in my experience) and view the change. You can also click on other areas of the image, where because of depth of field or lens softness at the edges, you can try another blur width , then navigate to a the least blurred spot and see what the higher blur width setting does there. You can then choose one or the other, or find a compromise in between and see the results immediately in FM's preview windows.
You can even make one or more broad selections, feathered, and use different blur widths for different parts of the image. I have found FM much more intuitive and accurate than Smart Sharpen. While the processing time for FM is longer than that of Smart Sharpen on my computer, the confidence with which I can use it is much greater than what I have experienced with Smart Sharpen.
Also, the reults of FM can be altered by the Fade control, either leaving it as is but selecting the Luminance blend mode, and/or reducing the effect of FM using the percentage slider.
Alternately, if you use FM on a layer above the background you can fade the opacity as well as the blend mode.
I use the 1DMkII, shoot RAW and convert to 16bit Tiffs in Prophoto RGB. I use Focus Magic as part of my workflow, and most often do not need to use other sharpening methods unless I uprezz. I do my uprezzing on the fly using Qimage and have printed up to 24x36 with Qimage's Smart Sharpening after interpolation by Qimage's Hybrid method. Using Qimage's Smart Sharpening at settings from 2 through 5 depending on the amount of small detail in the image, piggybacked onto the use of Focus Magic on the original processed file, produces prints that draw comments on the sharpness, with no artifacts of sharpening visible to the naked eye. I suspect if you used a loupe you could find artifacts but I put a sign over my images in exhibits saying "No Loupes." (Just kidding, of course.)
I print on an HP Z3100, and Qimage gives me more intuitive control over the prints and layouts than CS3 does. I have not printed from Photoshop for many years, thanks to Qimage. I do adjusting of the softproof duplicate in CS3 if needed, and send that to Qimage.
-
... the trick of using deconvolution methods is to determine the point spread function that is used in the deconvolution to undo the blurring. In the case of motion blur, this can be relatively simple. In the case of defocus, one method recommended by Focus Magic (http://www.focusmagic.com/) is to examine a point specular highlight and determine the diameter of the blur circle in pixels and use this for the deconvolution. I'm not sure how you would do this with Smart Sharpen.
- Am I correct that that Motion Blur's Remove option otherwise essentially is the same deconvolution as the Lens Blur option, except that the latter's effect is radial-equi-omnidirectional (to try compensating for bokeh), and the former's is linearly biased solely in the direction of the chosen angle?
- Using the Focus Magic estimation procedure, couldn't one zoom-in on a found small specular in an image, count the number of visually apparent bokeh pixels radiating around it, and insert that value in the Radius field? In other words, isn't Smart Sharpen's roughly-judged point spread function value supposed to be entered into the Radius field?
- I discovered another procedure I've never seen mentioned elsewhere, that may more accurately converge upon the optimum setting for point spread function (assuming I'm right that it's entered as the Radius value):
In this illustrative example, for purposes of clarity, choose a low-res JPEG, from which Smart Sharpen will bring out ugly artifacts. Moreover, choose the max 500% Amount, which of course can be dialed-back later.
Also, to accentuate things for this purpose, zoom-in to, say, 200% (notwithstanding that more realistic evaluation-guesses of regular USM on a computer screen, that we all know overstates any kind of sharpening - both accutance & bokeh-reduction, better might be viewed at, for example, 50% if going to press).
1) Using Lens Blur, observe that at a clearly too-large value of Radius for this particular image, such as 1.0 pixel, the noise & artifacts become worse when the More Accurate checkbox is enabled. Makes sense, considering that doing so applies the deconvolution twice.
2) But decrease the Radius value in 0.1 pixel increments, each time toggling the More Accurate checkbox, and at some point something very interesting happens: Below a certain Radius, the More Accurate setting drastically *reduces* the artifacts, yet the image still looks more focused.
I suspect that at this Radius point, the image's native point spread function has been reached!
How about you all giving this procedure a try, and let us know your findings back here?
... Bruce Fraser's work flow uses the unsharp filter with a radius determined by the camera resolution ...
Bless his soul, poor Bruce as far as I know apparently never overcame his propensities toward what Ken Rockwell characterizes as a "Measurebator": While image resolution certainly enters into it, good retouchers know that for any given image resolution, the particular fineness of detail in the image *content* significant area of interest is the paramount determinant of Radius. Whether in so-called capture, creative, or output sharpening, one would be a damn fool to apply the same Radius to, say, a glass of beer or wine effervescence as to furniture fabric with subtle intricacy. When the desired appearance of the subjectmatter detail sharpness is attained, the image resolution automatically has been taken into account along with it, without even thinking about it, as a matter of course.
-
Bless his soul, poor Bruce as far as I know apparently never overcame his propensities toward what Ken Rockwell characterizes as a "Measurebator": While image resolution certainly enters into it, good retouchers know that for any given image resolution, the particular fineness of detail in the image *content* significant area of interest is the paramount determinant of Radius. Whether in so-called capture, creative, or output sharpening, one would be a damn fool to apply the same Radius to, say, a glass of beer or wine effervescence as to furniture fabric with subtle intricacy. When the desired appearance of the subjectmatter detail sharpness is attained, the image resolution automatically has been taken into account along with it, without even thinking about it, as a matter of course.
[a href=\"index.php?act=findpost&pid=169975\"][{POST_SNAPBACK}][/a]
Ken Rockwell doesn't bother to use raw or a tripod and is generally laughed at in these forums, even though he does make some good points. If you read Bruce's book you would see that he combines sharpening for source and content into capture sharpening. For his rationale of sharpening for image source refer to p. 165 of his sharpening book, where his reasoning is convincing to me. Sharpening for image content comes next in the work flow and it does take the image characteristics into consideration.
In Bruce's output sharpening, the image content is irrelevant and he states that "creativity at this stage is not only unnecessary but should be discouraged." The determining factors are the final output resolution (in terms of PPI on the paper), paper type (glossy, matte, etc) and the type of output device, with contone printers needing less effect than an inkjet error diffusion printer or halftone process.
IMHO, it was Bruce's "Measurebation" combined with logical analysis that took into account the properties of human vision that set his work apart from guys like Rockwell.
Bill
-
I didn't hear anyone mention High Pass Filter method of sharpening. Why not? It's suppose to overcome the sharpening + increased noise problem. It might be me, but It also seems to give an image a "sheen" like melted glass smoothness.
-
It might be me, but It also seems to give an image a "sheen" like melted glass smoothness.
That's precisely what I don't like about HPF sharpening.
-
That's precisely what I don't like about HPF sharpening.
[a href=\"index.php?act=findpost&pid=177738\"][{POST_SNAPBACK}][/a]
So it's not just me. It does indeed give a sort of liquid look to the image. I'm not so sure I'm down on that look however.
-
When I first submitted to Alamy I had a D100 and a D70. Due to the smallish native files, I used GF V3 and later V4 to uprez. I also bought and used Photozoom Pro 2, and a whole host of software targeted at those who just wished they had bigger files. Later when PS CS2 came along, upsizing directly in one go was as good as GF, at least to my eyes.
Recently when prices became more than reasonable, I bought a demo D2X and followed by a new Canon 5D. Now to upsize those files to meet Alamy's requirements, I only have to increase size slightly, less than 30%. I kept thinking to myself, if I'm selling images and getting paid, then why am I being such a miser, so I went out and bought a Canon 1Ds MKIII and now have native size files larger than what Alamy is asking for. I'm done with upsizing because my task must surely be to worry about competing for sales with the few million other images on agency sites than to spend time tricking QC to accept borderline images.
I would seriously give it thought about looking at the whole process as a business that has to begin with the best quality capture one can achieve.
Just my 2¢
Henry