-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
Sharpening Images in Photoshop
- Thread starter LittleJim
- Start date
J
JGeddes
Guest
David: I asked the same question. It really depends on what other functions you use but it seems that out of the camera this works well.
100+ - - 3.0 - 10
James
100+ - - 3.0 - 10
James
I assume you mean functions of Photoshop. I only use the basics. Colour correct (Levels or AutoLeveling, which every looks better on a particular shot). and Brightness/Contrast.
David
David
A
Al Pacheco
Guest
I've been having good luck with the following:
In RGB,
Unsharp Mask....
Amount 200%
Radius 0.3
Threshold 2
Select Fade Unsharp Mask from the Filter menu.
Set Opacity to 100% and Mode to Luminosity.
This will apply the USM to only the luminance in the image and avoids the RGB> LAB> RGB conversion which is said to result in a loss of tonal gradation.
Keep in mind that the lens used to take the image is a factor, also.
I notice that when using my 50mm f1.4 that I don't need as much sharpening as when I use my 28-105 zoom. (kinda makes sense)
A lot of learning to do yet.
In RGB,
Unsharp Mask....
Amount 200%
Radius 0.3
Threshold 2
Select Fade Unsharp Mask from the Filter menu.
Set Opacity to 100% and Mode to Luminosity.
This will apply the USM to only the luminance in the image and avoids the RGB> LAB> RGB conversion which is said to result in a loss of tonal gradation.
Keep in mind that the lens used to take the image is a factor, also.
I notice that when using my 50mm f1.4 that I don't need as much sharpening as when I use my 28-105 zoom. (kinda makes sense)
A lot of learning to do yet.
You have me confused. Your first post asked for USM parameters, and after JGeddes gave you some good recommendations, you say you only use Levels etc. Anyway, there is an excellent discussion on sharpening in
http://www.dpreview.com/forums/read.asp?forum=1006&message=548595
The thread starts out with a new method using Genuine Fractals but includes good recommendations on sharpening with USM.
Hope it is useful,
Dave
I'm pretty much a Photoshop neophite, so I don't use much functionality. In general, I'll input a file, and colour correct it with Levels/AutoLevels, whichever gives the better results. Sometimes, I'll either change the brightness/contrast if needed, or use Curves if I only want the low-end Brightness adjusting.
I tried the settings you suggested, I assumed you meant
Amount : 100%
Radius : 3.0
Threshhold : 10
And it looked horrible! I'm guessing I did something wrong
David
I tried the settings you suggested, I assumed you meant
Amount : 100%
Radius : 3.0
Threshhold : 10
And it looked horrible! I'm guessing I did something wrong
David
I often have me confused as well!
JGeddes said "... depends on what other functions you use". I assumed he meant what other Photoshop functions are ALSO used, not used in in place of USM.
David
JGeddes said "... depends on what other functions you use". I assumed he meant what other Photoshop functions are ALSO used, not used in in place of USM.
David
B
Bill(y)
Guest
David,
parameters will change greatly according to each image, the size of the image (or area to be sharpened), the relative size of the detail to be sharpened, and the intended output and viewing distance.
The affect works by adding a highlite or halo between areas of significant color or value differences -- to accent edges or details -- and you don't want this to appear too stark, unnatural.
A 0.3 radius and a much higher percentage (like +250) would be a more appropriate place to start for photorealism. For prints on paper smaller than 11 x 14, I would try to keep the radius under 1.0.
Here's a non-abrassive place to start:
Amount = 100 (describes the sensitivity of the affect)
Radius = 0.7 (describes the size or width of the affect, the highlite or halo)
Thresshold = 5 (defines the minimum difference in value between pixels that should be targeted)
Be sure to look at your image at 100% in size. Start with the above parameters.
(Note: many prefer to sharpen an image in the L channel in "Lab" mode)
Play with the Amount. Amount is probably the more subtle of the three and as such has a wider range of applicable values; sometimes below 50%, other times all the way up to 500%. Amount tends to vary inversely to Radius.
Radius is much more sensitive. A finer number (a thinner highlite or halo) is better suited to images with fine details (twigs, hair, eyelashes) A Radius of .2 is often imperceptible. A Radius of .3 thru .6 is still relatively fine. 1.0 is getting big by most standards. Radius above 2.0 is colossal.
A Thresshold of 0 is the most sensitive. This will affect even the most subtle difference of color or value. This is appropriate to bring out surface textures, pollen dust on a flower.... Such a low Thresshold can also misinterpret noise, jpg or other messy artifacts, inadvertently highliting and drawing attention to areas which are better left smooth. A Thresshold near 20 will limit the highlite affect to the relatively larger object areas, areas that are already clearly defined by a distinct change in value or color.
Play with these sliders and note that what is appropriate for one section of an image may not be best for another area. Spot edits may be more appropriate than edge to edge applications of unsharp mask.
Also, other variables worth considering are the intended output media (paper, screen, canvas, t-shirt!) the size & distance of expected viewing (1 foot vs. 10 ft!) and, again, the relative size of the detail you are trying to sharpen (fine vs. large).
Hope this helps a little. Bill
parameters will change greatly according to each image, the size of the image (or area to be sharpened), the relative size of the detail to be sharpened, and the intended output and viewing distance.
The affect works by adding a highlite or halo between areas of significant color or value differences -- to accent edges or details -- and you don't want this to appear too stark, unnatural.
I really think 3.0 is way way way too large a radius, unless you're printing a poster or on a t-shirt....
A 0.3 radius and a much higher percentage (like +250) would be a more appropriate place to start for photorealism. For prints on paper smaller than 11 x 14, I would try to keep the radius under 1.0.
Here's a non-abrassive place to start:
Amount = 100 (describes the sensitivity of the affect)
Radius = 0.7 (describes the size or width of the affect, the highlite or halo)
Thresshold = 5 (defines the minimum difference in value between pixels that should be targeted)
Be sure to look at your image at 100% in size. Start with the above parameters.
(Note: many prefer to sharpen an image in the L channel in "Lab" mode)
Play with the Amount. Amount is probably the more subtle of the three and as such has a wider range of applicable values; sometimes below 50%, other times all the way up to 500%. Amount tends to vary inversely to Radius.
Radius is much more sensitive. A finer number (a thinner highlite or halo) is better suited to images with fine details (twigs, hair, eyelashes) A Radius of .2 is often imperceptible. A Radius of .3 thru .6 is still relatively fine. 1.0 is getting big by most standards. Radius above 2.0 is colossal.
A Thresshold of 0 is the most sensitive. This will affect even the most subtle difference of color or value. This is appropriate to bring out surface textures, pollen dust on a flower.... Such a low Thresshold can also misinterpret noise, jpg or other messy artifacts, inadvertently highliting and drawing attention to areas which are better left smooth. A Thresshold near 20 will limit the highlite affect to the relatively larger object areas, areas that are already clearly defined by a distinct change in value or color.
Play with these sliders and note that what is appropriate for one section of an image may not be best for another area. Spot edits may be more appropriate than edge to edge applications of unsharp mask.
Also, other variables worth considering are the intended output media (paper, screen, canvas, t-shirt!) the size & distance of expected viewing (1 foot vs. 10 ft!) and, again, the relative size of the detail you are trying to sharpen (fine vs. large).
Hope this helps a little. Bill
B
Bill(y)
Guest
I don't mean to belabor this....
I know the names imply as much, but 'Fading' an affect and selecting Luminosity isn't the same as applying the affect to the L channel of an image in Lab mode. Fading & Liminosity is a blending affect; it 'creates a resulting color with the hue and saturation of the base color and the luminance of the blend color. This mode creates an inverse effect from that of the Color mode. Bill
David,
Here's some great stuff to try - download Ultra-Sharpen at http://www.ickybits.com . It's free, I read about it on the Adobe forums, and consists of an action and automated plugin. Very nice results.
- Mohit
MichaelT
Veteran Member
Bill,
Very nice discourse on USM. I found it helpful. Thanks.
MichaelT
,Very nice discourse on USM. I found it helpful. Thanks.
MichaelT
parameters will change greatly according to each image, the size of
the image (or area to be sharpened), the relative size of the
detail to be sharpened, and the intended output and viewing
distance.
The affect works by adding a highlite or halo between areas of
significant color or value differences -- to accent edges or
details -- and you don't want this to appear too stark, unnatural.
I really think 3.0 is way way way too large a radius, unless you're
printing a poster or on a t-shirt....
A 0.3 radius and a much higher percentage (like +250) would be a
more appropriate place to start for photorealism. For prints on
paper smaller than 11 x 14, I would try to keep the radius under
1.0.
Here's a non-abrassive place to start:
Amount = 100 (describes the sensitivity of the affect)
Radius = 0.7 (describes the size or width of the affect, the
highlite or halo)
Thresshold = 5 (defines the minimum difference in value between
pixels that should be targeted)
Be sure to look at your image at 100% in size. Start with the above
parameters.
(Note: many prefer to sharpen an image in the L channel in "Lab"
mode)
Play with the Amount. Amount is probably the more subtle of the
three and as such has a wider range of applicable values; sometimes
below 50%, other times all the way up to 500%. Amount tends to vary
inversely to Radius.
Radius is much more sensitive. A finer number (a thinner highlite
or halo) is better suited to images with fine details (twigs, hair,
eyelashes) A Radius of .2 is often imperceptible. A Radius of .3
thru .6 is still relatively fine. 1.0 is getting big by most
standards. Radius above 2.0 is colossal.
A Thresshold of 0 is the most sensitive. This will affect even the
most subtle difference of color or value. This is appropriate to
bring out surface textures, pollen dust on a flower.... Such a low
Thresshold can also misinterpret noise, jpg or other messy
artifacts, inadvertently highliting and drawing attention to areas
which are better left smooth. A Thresshold near 20 will limit the
highlite affect to the relatively larger object areas, areas that
are already clearly defined by a distinct change in value or color.
Play with these sliders and note that what is appropriate for one
section of an image may not be best for another area. Spot edits
may be more appropriate than edge to edge applications of unsharp
mask.
Also, other variables worth considering are the intended output
media (paper, screen, canvas, t-shirt!) the size & distance of
expected viewing (1 foot vs. 10 ft!) and, again, the relative size
of the detail you are trying to sharpen (fine vs. large).
Hope this helps a little. Bill
...for Windows. Is there a Mac equivalent?
Karl Guttag
Veteran Member
I think we can all agree that a lot depends on image content, ISO/Noise content and what you want to bring out in the image.
My experience so far with the D30 agrees with JGeddes with the Radius of 3, 100% and levels 10. I find this combination sharpens edges without bringing out noise. When I tried a smaller radius and higher percentage and lower levels, the pictures start looking noisy before I get the edge sharpness I desire.
Upping the levels keeps a small mount of noise from getting amplified. The larger radius makes the radius cover the bayer color filter pattern (although the color conversion may do some of this too).
Now with my point and shoot 3MEG if I did a 100% radius 3 I would get ringing and noise all over the place, but the D30 usually looks about right. The D30 images, as most people have noticed, can really be pushed/sharpened without looking bad.
There is a good argument to go to LAB color and only sharpen in the Luma space. Theoretically this may be better but it depends on how much trouble you want to go through for a given photograph.
Note that a lot depends on the output size and personal preference. I have used Nik Sharpen software that is suppose to take some of these factors into account and it looks very over-sharpened on the view in Photoshop. I think they over-sharpen for smaller output images to get the details to come through.
I don't disagree with Billy's comments or reasoning, but with the D30 it seems to use a different set of parameter. I came to the exact same conclusion as JGeddes when I was looking at a Fuji S1 comparison to a D30 that was referenced in the news of DPreview. At first it looked like the S1 was sharper, but I found in Photoshop that I could sharpen the D30 to the same apparent sharpness as the S1 without increasing noise by increasing the radius (and I found that the S1 image could not be sharpened any more without significant ringing and/or noise).
So it looks like we are largely in agreement except on radically different default sharpening values :--). Looks like we will have Big Radius versus Small Radius camps. I was in the small radius camp before the D30, but I will be experimenting some more myself.
Karl
My experience so far with the D30 agrees with JGeddes with the Radius of 3, 100% and levels 10. I find this combination sharpens edges without bringing out noise. When I tried a smaller radius and higher percentage and lower levels, the pictures start looking noisy before I get the edge sharpness I desire.
Upping the levels keeps a small mount of noise from getting amplified. The larger radius makes the radius cover the bayer color filter pattern (although the color conversion may do some of this too).
Now with my point and shoot 3MEG if I did a 100% radius 3 I would get ringing and noise all over the place, but the D30 usually looks about right. The D30 images, as most people have noticed, can really be pushed/sharpened without looking bad.
There is a good argument to go to LAB color and only sharpen in the Luma space. Theoretically this may be better but it depends on how much trouble you want to go through for a given photograph.
Note that a lot depends on the output size and personal preference. I have used Nik Sharpen software that is suppose to take some of these factors into account and it looks very over-sharpened on the view in Photoshop. I think they over-sharpen for smaller output images to get the details to come through.
I don't disagree with Billy's comments or reasoning, but with the D30 it seems to use a different set of parameter. I came to the exact same conclusion as JGeddes when I was looking at a Fuji S1 comparison to a D30 that was referenced in the news of DPreview. At first it looked like the S1 was sharper, but I found in Photoshop that I could sharpen the D30 to the same apparent sharpness as the S1 without increasing noise by increasing the radius (and I found that the S1 image could not be sharpened any more without significant ringing and/or noise).
So it looks like we are largely in agreement except on radically different default sharpening values :--). Looks like we will have Big Radius versus Small Radius camps. I was in the small radius camp before the D30, but I will be experimenting some more myself.
Karl
Mike Kelley
Veteran Member
- Messages
- 1,445
- Reaction score
- 0
As long as we're talking about one of my favorite topics, let me once again point out what I still consider to be the best sharpening technique -- LAB sharpening.
Just change the mode to LAB, select the L channel (called Lightness, which is a misnomer in PS) and apply your USM only to that channel (if you check mark the top channel you can apply your USM while you view the effect in full color. Just don't select it by mistake). Convert back to RGB to do your levels and other adjustments.
LAB will not introduce noise into the color channels, and 95% of all detail is in the L channel. You can automate the whole thing with an Action (for one keypress convenience).
A
Al Pacheco
Guest
I can't claim any large degrees of experience in this matter. The actual method is from a photojournalist, Rob Galbraith. He states that converting to LAB and then back to RGB results in a loss of image quality. I'm sure there are many opinions on this. The fade Unsharp Mask is supposed to work on the unsharpened image in memory and is applied to the luminance only.
The fade in RGB is supposed to produce smoother results in out-of-focus backgrounds without the RGB> LAB> RGB conversion.
The fade in RGB is supposed to produce smoother results in out-of-focus backgrounds without the RGB> LAB> RGB conversion.
I don't mean to belabor this....
I know the names imply as much, but 'Fading' an affect and
selecting Luminosity isn't the same as applying the affect to the L
channel of an image in Lab mode. Fading & Liminosity is a blending
affect; it 'creates a resulting color with the hue and saturation
of the base color and the luminance of the blend color. This mode
creates an inverse effect from that of the Color mode. Bill
M
Matt
Guest
For us just learning folks, what is the proceedure to get your technique into
an "action" so that we can just click one thing and the whole process is done.
Thanks,
Matt
an "action" so that we can just click one thing and the whole process is done.
Thanks,
Matt
D
Daniel Lauring
Guest
It is obvious when you do the RGB to LAB and back the Photoshop is doing a serious conversion because of the time it takes. Anytime you do a conversion there is the possibility of lost data...the question is, how much. Have you noticed a loss of image quality yourself? Would it get worse if you were to run the conversion over and over?
Do you have a link to Rob's comments on this?
Thanks,
Danny
Do you have a link to Rob's comments on this?
Thanks,
Danny
The fade in RGB is supposed to produce smoother results in
out-of-focus backgrounds without the RGB> LAB> RGB conversion.
I don't mean to belabor this....
I know the names imply as much, but 'Fading' an affect and
selecting Luminosity isn't the same as applying the affect to the L
channel of an image in Lab mode. Fading & Liminosity is a blending
affect; it 'creates a resulting color with the hue and saturation
of the base color and the luminance of the blend color. This mode
creates an inverse effect from that of the Color mode. Bill
A
Al Pacheco
Guest
I'm afraid Rob doesn't post his observations online. He wrote a book,
http://www.robgalbraith.com/digiguide/index.html
which has a chapter dedicated to Photoshop (version 4-5).
He mentions the heavy math that PS performs in the conversion process and how that would result in some errors.
http://www.robgalbraith.com/digiguide/index.html
which has a chapter dedicated to Photoshop (version 4-5).
He mentions the heavy math that PS performs in the conversion process and how that would result in some errors.
Do you have a link to Rob's comments on this?
Thanks,
Danny
The fade in RGB is supposed to produce smoother results in
out-of-focus backgrounds without the RGB> LAB> RGB conversion.
I don't mean to belabor this....
I know the names imply as much, but 'Fading' an affect and
selecting Luminosity isn't the same as applying the affect to the L
channel of an image in Lab mode. Fading & Liminosity is a blending
affect; it 'creates a resulting color with the hue and saturation
of the base color and the luminance of the blend color. This mode
creates an inverse effect from that of the Color mode. Bill
D
Daniel Lauring
Guest
Thanks for the link. While I imagine this book would be great, I think I want to wait for the next edition where he will most likely address the D1, D30 and Photoshop 6. I know he says not to count on that but I'm hoping he gets the itch. I have been very impressed by Rob's reviews and website.
He mentions the heavy math that PS performs in the conversion
process and how that would result in some errors.
Do you have a link to Rob's comments on this?
Thanks,
Danny
The fade in RGB is supposed to produce smoother results in
out-of-focus backgrounds without the RGB> LAB> RGB conversion.
I don't mean to belabor this....
I know the names imply as much, but 'Fading' an affect and
selecting Luminosity isn't the same as applying the affect to the L
channel of an image in Lab mode. Fading & Liminosity is a blending
affect; it 'creates a resulting color with the hue and saturation
of the base color and the luminance of the blend color. This mode
creates an inverse effect from that of the Color mode. Bill
P
pdunnuck
Guest
How would the action step work if each photo is different...takes different sharpening?
Also, do you beleive the GF step is worth the trouble...GF... to tif then sharpen...back to GF? What is the point of the tif step? Why not sharpen in the GF mode and eliminate the tif step?
Also, do you beleive the GF step is worth the trouble...GF... to tif then sharpen...back to GF? What is the point of the tif step? Why not sharpen in the GF mode and eliminate the tif step?
Keyboard shortcuts
- f
- Forum