Want Tack Sharp...please help

Barefoot and Natural

Hi everyone,
I am trying to get the below image crystal clear!! I don't know if it can be done because it wasn't super sharp to begin with!
I need this picture perfect.....this a REAL GENUINE SMILE from my daughter and this is rare!! She is my serious gal and I can hardly ever get a real smile out of her!!

I shot it in RAW, but my software is what came with my 20D and that is what I used for this.




I am still trying to find out if I can put a RAW image on here so you can see the original!

Thanks everyone and anyone who can help!
Heather

jdryan3

I don't see an image. Just the red X.

pathfinder

I don't see it either.

Do you have external links turned off for this gallery?

Barefoot and Natural

pathfinder wrote:

I don't see it either.

Do you have external links turned off for this gallery?

Sorry, it should be there now.

schmoo

Image is still not showing. My curiosity is going crazy now.

Barefoot and Natural

schmooo wrote:

Image is still not showing. My curiosity is going crazy now.

Ok, I must be doing something wrong, I am going to try it again.

Barefoot and Natural

Barefoot and Natural wrote:

Ok, I must be doing something wrong, I am going to try it again.

How 'bout now?

pyrtek

Not much one can do with such a small file, but still, a little USM improves it.

Duffy Pratt

I agree that the file size presents some pretty severe limitation.

The basic sharpening strategy I tries was to do a regular USM line sharpening on the K channel in CMYK, and a hiraloam sharpen on the C channel. The idea was to leave the skin tones alone, sharpen the hair and eyes, and use the hiraloam sharpen to add some interest in the water.

Before I did that, however, I decided to see if I could mitigate some of the blue light reflecting on her, particularly under her upper lip. I converted to LAB for this, and applied a warming filter using an inverted B channel both blurred and steeply curved to emphasize the effect where the image is bluest. This had the nice effect of adding texture to the pool, and took away some of the nasty blue reflection that was bothering me.

Then I did Margulis standard trick of overlaying the A and B channels on themselves, reducing opacity to about 17%. That got the colors about where I wanted them. Then I moved to CMYK for the sharpening. (I didn't like the L channel sharpening. If I was going to take more care on this, I would probably have made a duplicate image to extract the K channel and use it inverted for a sharpening mask. But this did the job quickly without too much damage.)

Duffy

Barefoot and Natural

Here is link to the orginial jpeg file:
http://barefootandnatural.smugmug.com/gallery/1610491/2/139709824#171584407-O-LB

I really thank you for this!!!

Barefoot and Natural

Love the instructions you gave! I am still trying to learn everything you can do to a RAW image and it is so mind boggling!!

I feel I might have overprocessed this image because her eyes are a hazel-green and you can't see that detail in the image I am showing.

nikos

A combination of techniques



The Original

Nikos

pathfinder

I think Duffy and Nikos both did a very nice job.

I agree that this image has a significant blue cast. It is not in focus, and probably suffers from camera shake as well...(.Not a nasty comment, I am just trying to say that there is a limit to what can be done with this file in my hands)

I downloaded the original jpg and opened it into Photoshop CS3 via the RAW converter (which is a nice feature of ARC 4.1) I decided to see if the light are along the bottom of the image was white or off white and used it with the eye dropper to set my initial white balance. I then opened into a jpg, and looked at the values for the skin on the forhead - I increased the red curve slightly to pull down the cyan cast, and I lowered the blue curve to increase the yellow slightly. I did us a sharpened K channel to blend with the RGB image and did a smart sharpen on the L channel as well.

After all this I still felt the skin was slightly too cyan, so I added a little more curve to the red channel, and decided to stop there.

This is my result



The full size file is available here -

http://pathfinder.smugmug.com/gallery/1789718/2/171737689/Original

pyrtek

As others have said - you will never get this picture "tack sharp". Here's my
attempt at a slight fix:





Change -3 to -5 in the URL for original size.

BinaryFx

I would be looking to reproduce this image at the smallest size possible - if lucky there would be way too many pixels for the output process/size and one could resample down in size and then try CS2 smart sharpen or just sharpen in an aggressive manner through an edge mask and with hiraloam method on the duped resampled down image.

I have some other ideas but no time to look deeper at the moment.

Good luck,

Stephen Marsh.
http://members.ozemail.com.au/~binaryfx/

Barefoot and Natural

I really want to thank all of you for trying!! I am sure most of you know that getting that perfect picture of you own children is sometimes a small defeat in itself!
I am happy to report that I was able to get ANOTHER great smile out of her yesturday! I guess it is the Summer Time happiness coming out!!!
Granted, this one has her BESTEST cousin in it, but I love it just the same!!




I just may convert that first one to b&w and play with it, so it looks as though the oof was intentional.


Thank you again for all your help!!
Heather

edgework

Couldn't get the original jpg to download, but I tried this with the image first posted.



Used a couple of high-pass layers, one at .5, one at 1, both in Linear Light mode at 100%. This created some godawful halos, but sharpend it nicely.

Merged to a new layer, turned off the high pass layers, duped the merge layer and used the old trick of putting one in darken mode, the other in lighten mode at 50%.

Made a group/layer set of those two layers and reduced the set to 80%.

A slight tweak in the red channel and a boost in yellow to the blue channel.

ziggy53

Since it looked like the image was truly OOF, I used "Image Analyzer" and its Deconvolution Filter, Gausian (default settings).

I felt this gave a pretty good rendition of the hair and good detail in the eye highlights. I did enlarge the catch light size in the eyes in PS.

I used PS to reduce the reflections that I felt were not necessary on the face.

Some USM and some exposure and color work and crop to 8 x 10 and a little cloning and I came up with:

BinaryFx

ziggy53 wrote:

Since it looked like the image was truly OOF, I used "Image Analyzer" and its Deconvolution Filter, Gausian (default settings).

Ziggy, I will look into this app as I am always interested in deconvolution to help restore camera or minor motion blur. I was expecting more from smart sharpen in CS2 in this regard, I was not expecting magic, but can hope for more in future releases. Another two options:

http://www.focusmagic.com/

http://www.fixerlabs.com/EN/photoshop_plugins/focusfixer.htm

I would be interested to just see what Image Analyzer's deconvolution does with no other edits. The image that you posted has lost a lot of contrast when assuming sRGB or common non colour managed web browser viewing and I am not sure if it was due to other edits or using default settings etc.

I have a three step method of attempting to restore the appearnce of minor focus to an out of focus image, but it does not work well with movement. It uses a large DoG sharpen (difference of Gaussians), similar to HiRaLoAm (high radius, low amount) but less sensitve to noise than unsharp masking. Then it uses an overlay blended emboss filtered version of the image to fool the eye into thinking that things are a little more in focus. Optionally as the last step, the lighting effects filter is used, with the L channel of LAB pasted into the RGB file as a bump map channel for the lighting effects filter (the idea being to affect brightness as little as possible and not to introduce vignetting type effects at the corners etc). This is luminosity blended at reduced opacity/blend if sliders.

The DoG band pass filter simulation and many other simple actions are included in an action set found here:

http://members.ozemail.com.au/~binaryfx/convolution.html

I did not use curves first, but one should have a good steep curve/contrast in the image tones of interest, which will make an image appear "sharper" with no sharpening. EDIT: One can also add 'midtone sharpening' methods in Photoshop or 'clarity' in ACR4.1 or ALR1.1.

Attached is a JPG crop at original size. Left is the unedited sRGB image originally posted. Right is the 3 step method outlined above - note that no USM filtering or Smart Sharpen filtering or deconvolution has been applied to achieve the sharper version to the right. I would hope that deconvolution software can do a better job than this very quick attempt in Photoshop CS.


Sincerely,

Stephen Marsh.
http://members.ozemail.com.au/~binaryfx

ziggy53

ziggy53 wrote:

Since it looked like the image was truly OOF, I used "Image Analyzer" and its Deconvolution Filter, Gausian (default settings).

I felt this gave a pretty good rendition of the hair and good detail in the eye highlights. I did enlarge the catch light size in the eyes in PS.

I used PS to reduce the reflections that I felt were not necessary on the face.

Some USM and some exposure and color work and crop to 8 x 10 and a little cloning and I came up with:

My apologies about the above image I posted. I processed this on my "Huey" calibrated system and I didn't check the image on any other system before posting. While it does look fine on the "Huey" system, it looks pretty bleached on my other systems.

I'll tackle it again when I have time. In the mean time, it should probably look more like this:

fehlinger

I love it when we're invited to fix a file! I'm a bit late to the party but I had a go at it. Changed the color cast, adjusted for a warmer skin tone (did I go too far?), used Dan M's LAB steepening of the A & B channel, brightened the eyes a touch, USM, Noise ninja (I know, its an ISO 100 shot, but I find it can clean up even low noise shots). Cute shot, by the way!

Cary

rwells

I might as well give this a whirl also

Took posted image from first page...

Duplicate layer -- USM (30-30-0) -- this sounds strange, but check it out - also gives pic some "pop"

Duplicate layer -- USM (100-1-0)

Duplicate layer -- Noise Ninja

Adjusted skin tone a bit

Adjust opacity of layers to taste...

BinaryFx

ziggy53 wrote:

My apologies about the above image I posted. I processed this on my "Huey" calibrated system and I didn't check the image on any other system before posting. While it does look fine on the "Huey" system, it looks pretty bleached on my other systems.

I'll tackle it again when I have time. In the mean time, it should probably look more like this:

Thanks for the notice ziggy, I was not sure if the lightening and loss of contrast was due to the deconvolution default settings or the other edits performed, so it would appear to be the latter, led astray by the monitor appearance. This is why many evaluate images both visually on colour managed systems (hopefully accurately setup) and "by the numbers" - be they RGB, LAB, CMYK, Grayscale etc. If the monitor appearance is saying one thing and the numbers another, then there may be a problem. Even with a poor Huey (or other make/model) calibration, the other images on the site from other users would have looked similar to each other for tone, while your image would have been very different (I am sure the Huey can do better, perhaps ambient light was a factor when editing?). This should in future be considered something to take note of. It can also be useful to have a printed reference image to compare to the monitor image to help establish calibration and colour management. Just as when printing, it can be helpful to compare a current print of a older print that is known to be accurate.

I have downloaded Image Analyzer and will check it out later tonight, thanks for heads up.


Best,

Stephen Marsh.
http://members.ozemail.com.au/~binaryfx