| Author | Thread |
|
|
01/26/2007 10:08:18 PM · #76 |
| Difficult subject those branches, like hair the sharpening can be overdone very easily especially when resized. |
|
|
|
01/26/2007 10:11:12 PM · #77 |
Trying to make the jump to print media rather than digital, I'm seeing noise everywhere. It haunts me...
|
|
|
|
01/26/2007 10:23:00 PM · #78 |
| Heh. Re-convereted the RAW file, and the noise is less significant, though the gentle mottling in the sky is visible. Sharpening and curves tweaks don't enhance it much at all, leading me to believe that Photomatix had a hand in that. Hmmm. |
|
|
|
01/26/2007 10:30:01 PM · #79 |
| OK, resizing from a non-tonemapped original, converted directly from RAW and kept in 16-bit still shows the "lumpiness" in the water, but abain a perfectly smooth sky. I'm convinced that the "texture" of the water in the resized version is real detail, but I think that my USM pass for local contrast enhancement got out of hand there, comparing the new conversion with the old. Lesson learned! |
|
|
|
01/26/2007 11:26:12 PM · #80 |
Originally posted by Gordon:
...To me, properly sharpened means that the sharpening is not visibly noticeable... |
Yes. And this can happen all to easily with patterned textures, when it results in moire-like effects, which can be either hell or heaven, depending on what you want or need -when it works!- with a particular image.
Because of the varying characteristics of the images I process, I like to first find out and note where, precisely, its limits are in terms of sharpening (as Gordon has it 'not visibly noticeable'). During a second pass I try to decide whether the image would benefit from recess or from a transgression beyond that limit. The final pass(es) are always deductive.
Funny thing is, I often sharpen (and undo) immediately after conversion from RAW, just to see... |
|
|
|
01/27/2007 12:45:09 AM · #81 |
I'm looking at the 7Mb file in CS2 and commenting as I go. My overall impression is that Kirbic is dealing with the uber-problem of wide angle landscapes on the web; the level of detail overwhelms the ability of the screen resolution to capture it. I run into this all the time. The thing you notice in Photoshop is that up to a point, the more you zoom in the sharper it looks. The problem is, how to emulate that sense of sharpness in DPC-size images without having it all break down into tiny little details that create distracting visual chaos.
1. Working on full-size image. I have duped the BG layer and applied a little shadow/highlight to the image. I have slightly mellowed the shadows and I have muted down the highlights a bit. It is a little flat right now because I have another step that will ramp it back up a little later.
2. I have selected foreground color in the color picker to be the darkest part of the sky. I have selected the sky with the magic wand and "select similar" and saved the selection. I have created a new empty layer in multiply mode and added a foreground-to-transparent gradient of the sky color, and faded that layer to a reasonable density. The sky looks more even now.
3. I have made a rectangular selection of the entire foreground and opened a hue/sat layer: I have desaturated and lightened both blue and cyan about -15 sat and +24 lightness: the water looks more natural now to me.
4. I have created a gothic glow layer, lowered the saturation on that, run aggressive high pass sharpening on the glow layer, copied it, and pasted it onto the image in progress. I have dropped the layer opacity of that 20 20%. The image now looks overall sharper than the original, but at the same time richer and softer somehow. We are getting there. No hint of oversharpening has yet shown up, and the image looks more elemental and less cluttered at normal viewing size.
5. On a new, multiply layer filled with a white mask I have laid in vignetting. I have faded that layer to best effect, slight vignetting contains the image somewhat.
6. Now I have merged all layers, copied the merged layer set, then reverted to the pre-merge state. I have pasted the merged layerset as a new layer, duplicated that, and applied 2 pixels of high pass to it in an overlay mode. We are now looking nice and sharp without any distracting oversharpening, but we are still at full size, of course: the test is yet to come. Incidentally, I merged and unmerged/pasted so I'd still have the complete layer set underneath the new composite layers, in case I want to fade down the composite. It doesn't appear that I will need to.
7. I have resized to 720 pixels, same as entry. No further sharpening seems to be required beyond the bicubic sharper algorithm of the resize, in this case. Blowing up the final image significantly shows trace haloing on the branches and a bit on the rocks, but not an excessive amount considering the constraints of the detail here. The overall effect is more evenly-toned to my eye, less likely to be called "oversharpened".
What do you think? I am doing this on a 22 inch CRT, calibrated, at 1600 pixel screen resolution.
(whew)
Robt.
|
|
|
|
01/27/2007 03:12:34 AM · #82 |
Originally posted by Gordon:
I had the interesting experience this evening when walking outside. It was a bright, but overcast day. Plenty of bare, thin twigged trees around.
Looking at them with an 'image processing' eye, they looked 'oversharpened' in real life. |
Gordon, that makes sense given the article you referenced earlier - our brains add contrast to edges to help us see them even when they're not there.
|
|
|
|
01/27/2007 10:08:51 AM · #83 |
If you process from RAW another question related to sharpening is this:
After flattening and resizing, should you sharpen before or after you convert back to 8-bit?
Before seems more logical to me but what do you think?
|
|
|
|
01/27/2007 12:55:35 PM · #84 |
Originally posted by stdavidson:
If you process from RAW another question related to sharpening is this:
After flattening and resizing, should you sharpen before or after you convert back to 8-bit?
Before seems more logical to me but what do you think? |
An interesting question to be sure. As a matter of being conservative, doing it before makes sense. I don't see a downside to this, except possibly it taking a fraction of a second longer ;-)
Doing it afterward could, I suppose, result in more artifact, but I'd bet that the difference would be extremely small.
FWIW, as a matter of practice, converting to 8-bit is the very last step I do prior to saving; as a matter of fact, at least in CS3, and I believe CS2, yoiu don't have to explicitly convert at all if "saving for web." The conversion is done as part of that process. |
|
|
|
01/27/2007 01:06:40 PM · #85 |
Originally posted by kirbic:
... FWIW, as a matter of practice, converting to 8-bit is the very last step I do prior to saving... |
Fritz, being the logicians and techy types that we are I'm not surprised we do things the same way. :)
But when I hear the simple and elegant way that idnic handles sharpening I can't help but wonder if I overthink the task. LOL!
Btw, I have an entry in the January Free Study that absolutely depends on sharpness for its overall effect and I have high hopes for it, so find this discussion both interesting and timely.
Note:
Though I find Bear_Music's treatment interesting I think your 3rd edit is the best treatment of your entry I've seen posted. You toned down the trees and that makes all the difference.
This discussion is particularly meaningful to me because I did a 16 X 20 print of Cathedral Rock for a gallery showing and the sharpness of the trees was my biggest issue!!!
Message edited by author 2007-01-27 13:15:53.
|
|
Home -
Challenges -
Community -
League -
Photos -
Cameras -
Lenses -
Learn -
Help -
Terms of Use -
Privacy -
Top ^
DPChallenge, and website content and design, Copyright © 2001-2025 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 09/13/2025 11:51:54 AM EDT.
