I love this cute picture of my son on a carousel, I don't love the blur. I have been playing with Photoshop for a long time now but haven't come close to mastering de-blurring. What can I do here (details please!)?

The Picture (http://www.luckeyfamily.com/mallcarousel-20mo-sm.jpg)

This is about all I could do, used a plugin called Focus Magic, adjusted levels on each channel and ran USM.

Hi KeepEmComin,

Unfortunately there really is no way to re-focus a badly out of focus picture. The sharpening tools of PS and other Image Editing software depend on edge emphasis, which gives an optical illusion appearing to sharpen the outline. However, this is only good for pictures that are only a little out of focus. Your image is too blurred to ever get a truly acceptable result.

I've had a go with your image, by creating an edge mask, applying USM, then applying local sharpening to eyes and mouth. Then I selected the boys face, copied to a new layer, applied USM again, and adjusted levels.

Result is not really acceptable, but it's as good as I can do.

Hi everybody,


keepemcomin

what a lovely little boy you have!!!!!

What Gary said about out of focus pictures and sharpening remedies.

You can still try to emphasize details as to improve it anyway... :wink:

Here is what I did with your picture:


Duplicated the background Layer.
Run Levels on each Channel to correct the colour and contrast.
Used the 'Curves' method as drscribed in this Tutorial (http://retouchpro.com/tutorials/?m=show&id=65) to bring out a bit more details in your little boy's face. :happy:
Merged Visible. (here (http://www.retouchpro.com/forums/showpost.php?p=67894&postcount=18) is how to do it without losing the underlying Layers)
Duplicated the merged Layer, run a Radial Blur (Zoom) on it. (just for a different blurring effect) ....
Added a Layer Mask to the blurred Layer and, with a soft black brush I painted over your son only to uncover the unblurred 'version' from the underlying Layer.
Merged Visible.
Used USM (Amount>20%, Radius>50, Threshold>0) to further enhance the contrast.
Created a Shadow Mask, (Ctrl+Alt+~= Luminosity Mask > Ctrl+I = Invert = Shadow Mask), copied it on its own Layer (Ctrl+J).
Used USM to sharpen the shadow Layer only ...


I know it sound like a long, complicated procedure ... but it only took ten minutes ... By blurring the 'background, enhancing colour and contrast and just a light final sharpening I wanted to make your son really 'pop out' in a natural way ...


vinniesworld,

Nice job! .... a tad oversharpened ... maybe?

Gary,

I think your version is very nice!!!

Hope this helps..

Thanks for your comments Flora. Your sharpening attempt is as good as I've seen. However, I still stick with my original comment, that you cannot truly sharpen an out of focus picture. This is not to say that no attempt should be made to improve it, but that the results may not be all you hoped for.

Gary,

I agree absolutely with you!

I haven't seen a method yet that can work miracles on out of focus images ...
That's why I try to sneak around the real problem by enhancing the main subject of the picture and lightly fading the rest .... but, as you said, not even close to a crystal clear new image which may be expected as final result...

Thank you all very much! I appreciate your pointers, I'll go try them out now and see if I can get some decent results.

I really like the idea of blurring the background to make my son stand out more. It's sort of a "fight fire with fire" approach - the picture is blurry, so blur it more. :)

Thanks again!

Gary,

I agree absolutely with you!

I haven't seen a method yet that can work miracles on out of focus images ...
That's why I try to sneak around the real problem by enhancing the main subject of the picture and lightly fading the rest .... but, as you said, not even close to a crystal clear new image which may be expected as final result...

Too bad there's no real software like they use on the various TV cop shows. I think it was CSI recently where they had a photo and were able to zoom in on a small pill bottle in the background that was way out of focus. Then then sharpened it up until it was perfectly clear and they could read the perscription information! :dizzy: Impossible image manipulation comes up like that all the time on TV shows and movies.

So does being able to input any instruction no matter how complicated with 2 or 3 keystrokes, the ability to guess any password in 2 or 3 tries, and the fact that everyone can be hacked through the internet in 2 minutes, even if they're offline. I don't think TV really has too good a handle on computers.

First I must register my apologies to keepemcomin for a thread hijack, but I just had to stick my big nose in here.
However, I still stick with my original comment, that you cannot truly sharpen an out of focus picture.....

Something to think about.......
If we stick to a strict mathematical sense, that is the pure truth. A definition of a blurred photo would be one in which the higher frequencies have been eliminated. Automatic (as opposed to manual) sharpening relies on amplifying these high frequencies. And so, it follows that as there is nothing there, we end up amplifying just noise.
But if we think of information, the story is quite different. Looking to the left of the photo, we see written "COIN SLOT ON OTHER SIDE" (looks like "SLOY" but we know what the word is). Given that, we can manually sharpen / redraw the text to exact perfect edges.

OK, so that's an extreme example. But look around the photo we could manually sharpen / redraw the pole, the horse's head, the whole merry-go-round...

OK, again, but what about the face?
Lots of people send on old, torn, faded, incomplete photos to the Restoration threads, and it doesn't take long before somebody posts a perfect reconditioned image. An ear missing? Copy from the other side. A nose? Copy from the other guy. Torn? Copy, stamp, heal from nearby. That information wasn't there, but we know what a face looks like so it's easy (well, possible) to fix.
Here's something to think about. If keepemcomin had posted a photo with a hole, a piece torn out or a bad coffee stain on the kid's face, the experts from the Restoration threads would have whipped up a repainted 'perfect' version in a matter of minutes. But, as it was only blurred, nobody felt free to push pixels around and just stuck with the automatic (tweakable) tools.

Like mbrady said. You cannot see things that simply aren't there. So we can't count every hair on the kid's head, but there is nothing here that we couldn't restore / repaint manually to full sharpness. (the bar code? OK, I'll give you that one).
Rô

Hi Ro,

I'll be more impressed with your answer when I see you do what you suggest on the whole picture.

Of course all things are possible theoretically. If, you had another picture of the boy, If, you had another picture of the mother, If, you had another picture of the carousel horse, and if all these other pictures were lit similarly to the image you wish to restore.

But this is not likely to be the case, and using someone else's face on the boy, is not likely to be un-noticed by the mother who knows her own sons face intimately.

But I'll be happy to be proved wrong, so go ahead and have a go, and please post your results.

Hi Gary,
As in all things in life, and in a good photo, there should not be too much black and white - but a whole lot of grays. As such, I was disagreeing to the idea that..
....you cannot truly sharpen an out of focus picture.
because that seems like putting up a brick wall. (OK, we could discuss "truly". Is a finite-pixel, finite-channel-level, jpeg-compressed, digital file a true representation of a nice kid's face?)
As to needing another picture, that depends. I did a quick fix on the text and there was no other picture. I could do a pretty good fix on the horse, because it's full of straight lines and plain colors - and because there would be no horse's mother to worry about.
And the face? Have you ever noticed how a bad photo on the screen looks much worse than the orignal in you hand. We all have a great ability to see through the 'noise' (in fact, adding noise is a good sharpening trick). If we look at the blurred face at 100 or 200% zoom it's real bad, but look at a smaller version and it's much clearer - so (all?) we need to make the zoomed version look like we see the small version.

Like you said....
Of course all things are possible theoretically....
but that doesn't imply that all things are impossible in practice (although it would take some time!).

Once again, apologies to keepemcomin for the hijack.

Rô

I'm a newbie to this and this is my first attempt to reply to threads like this.

I use a sharpening program designed by Bruce Fraser called PhotoKit Sharpener. I'm still learning to use it but it works wonderfully for work I have at the studio.

It has three levels of sharpening and does it all on layers. Capture sharpening, Creative and Output. Not sure I used the correct ones for this photo but thought I'd give it a try.

Vicki

Ro,

I'm not in disagreement with your ideas in principal, merely in the application of them in practise. What you are suggesting may be possible (I still hold some reservations on that), but until I see an example, I will remain sceptical.

vidanse,

Looks OK, but hard to tell, your posted image is too small to judge. (This proves Ro's point about things looking sharper the smaller they are.)

Created a Shadow Mask, (Ctrl+Alt+~= Luminosity Mask > Ctrl+I = Invert = Shadow Mask), copied it on its own Layer (Ctrl+J

Flora, could you elaborate on this??
When I copied the invert to its own layer, it was kind of a funky color. Correct? Am I supposed to change the blend mode??
Do I add a mask, or is the selection itself a 'mask'?
This one's new to me, can you tell?
Thanks
kiska

NEVERMIND! On my mac, to invert a selection, it's SHIFT+command+I. Command(ctrl)+I inverts the whole image. That's the funky colors.Thanks.

I'm not in disagreement with your ideas in principal, merely in the application of them in practise. What you are suggesting may be possible (I still hold some reservations on that), but until I see an example, I will remain sceptical.

Hi Gary,
I agree. Although I it seems I'm just a little less sceptical!

Hi kiska,
I know you asked Flora :wavey:, but I'll chip in here:
When you press Ctrl+Alt+~ you are actually making a selection. When you make a selection manually, you see things as being fully selected or not selected at all - as shown by the 'marching ants'. However, in this selection the lighter parts (more luminosity) are fully selected and the darker as less selected. It may seem that the 'ants' have lost their way!
The Ctrl+I is a command to invert the selection. So now the dark parts are more selected and the light parts less so.
Now the Ctrl+J is the normal duplicate layer command. Which, (as most people first discover by accident!) will copy only the selected parts to a new layer.
Putting that all together, you should now have a new layer showing the darker parts of the original and transparency where there were some lighter parts.

OK, now?

Rô

Thanks Ro, I understand selecting luminosity, invert, copy.Flora called this a 'mask'. Is that it, or do you add a mask to it. Do you leave the blend mode normal or not??
kiska

AHA! To invert the selection, on a mac, it's SHIFT+command+I. I was omitting the SHIFT, therefore funky colors as the whole image was inverted. Sorry.

AHA! To invert the selection, on a mac, it's SHIFT+command+I. I was omitting the SHIFT, therefore funky colors as the whole image was inverted. Sorry.
:o: Ooops! :o: (I just copied without thinking, I'm sorry)
It's not a PC / Mac thing - to invert a selection on a PC it is Shift + Ctrl + I .
I understand selecting luminosity, invert, copy. Flora called this a 'mask'. Is that it, or do you add a mask to it. Do you leave the blend mode normal or not??

The idea behind a selection and a mask is the same. However a mask is tied to a layer, but a selection applies anywhere until de-selected.
To illustrate (or to confuse completely) try this:
- Without anything select (command + D) duplicate the photo (command + J);
- On the duplicated layer make the luminosity selection (command + alt + ~) and invert the selection (shift + command + I - got it right now!);
- Click on the "add layer mask" button at the bottom of the channels palette;
- Turn off the original and you'll see the same result as you had using Flora's method, but now using a mask and not a copied selection.
(after this you can use "apply mask" and you arrive at exactly the same place as the other method - but now by the scenic route)

Usually Flora's method is better because when you are working on the "shadow" layer there is no possibility of some nearby lighter bits, that were hidden by the mask, messing things up.

As to blending, usually you'll just leave as normal - maybe lowering the opacity if you exagerated on the touch-up. Of course all the options are open.

Rô

I was wondering about the blend modes because I was getting funky colors when inverting. Now that i am SHIFTING SHIFTING SHIFTING the mode question is moot.
Thanks for comparing the shadow mask to a 'mask' mask. It helped the light to dawn.
kiska

Here is my try :)

Tried a win software called Cadet used for de-blurring (and other processing) of astronomical photographs. Selected a blurred point to represent the PSF and used Maximum entropy deconvolution. Other settings were left as such since I do not understand much of these. Got a reasonable sharper picture, if not the best.