|
The
FFT (Fast Fourier Transform) technique offers a way to remove (most of)
the texture, leaving us with a photograph that can now be easily
retouched.
Here’s a before / after image to show what we’re talking about….
On the left we have an image (thank you, FrannyMae) scanned (*1) from a photograph on embossed paper – on the right the result after applying the technique described here.
Some observations:
The FFT technique...
...has eliminated the superimposed embossed texture;
...has preserved the underlying features;
...has NOT removed spots and random noise;
...has NOT fixed any colour blotches.
What is FFT?
Without
going into too many gory details, FFT is a mathematical transformation
that takes a “normal” image (that we can understand) and transforms it
into an alternative “mathematical” image (that only a computer
understands).
What is the FFT technique?
Using
the “mathematical” image we can easily identify and eliminate the
repetitive texture. When we transform back (called, Inverse FFT) to the
“normal” image, the texture has been removed.
Preparation
You will (of course) need an FFT filter. There are some options (*2) available, but the filter that has been most widely adopted is the version by Alex Chirakov. (*3 download and installation).
The method described here is for this filter but, with some minor adaptations, can be applied to all other FFT implementations.
How to apply the FFT technique:
1) Duplicate original image
- Right click on title bar;
- Duplicate (merged, if it has more than one layer) to a new image.
2) Flatten all Layers
You may not need this but, for precaution, flatten all the layers...<alt><L>, <F>.
This is because FFT doesn’t understand layers, masks, alpha channels
etc. The image must be just one “Background” layer and nothing else;
3) Run Filter>Fourier Transform>FFT RGB;
You should be looking at something like this…
(don’t worry!)
4) Select just the Red Channel;
· Go to the channels palette
· Select the Red channel only (or turn off the Green/Blue)
· You now have something like this
(don’t worry!!)
Do
you see the little stars? These small white points correspond to the
texture that is superimposed on the photo. All the rest, especially the
central part, is the image that we want to preserve.
5) Paint out stars in the Red Channel
- Using a soft brush paint the stars out with a dark grey or black. .
- DO NOT TOUCH THE CENTRAL STAR
- You should have something like this..
(don’t worry!!!)
6) Select all channels again
- On the channels palette turn back on the green and blue channel (easy to forget this bit);
- If necessary, flatten the layers (<Alt><L><F>);
7) Do the Inverse Fourier Transform
- Run Filter>Fourier Transform>IFFT RGB;
- What no colours? That’s right! The FFT works only on the luminosity (greyscale) information.
8) Put the colours back in:
- Slide, with <Shift> pressed (to align), your (now un-textured) luminosity back layer on top of the original (that’s the one we left behind in step 1);
- Set this new layer's blending mode to Luminosity.
- Select your original layer, which is below the new Luminosity layer;
- Apply a small radius (2 or 3 should do) Noise>Median filter (this blurs the colours a little, but don't worry all the important sharpness is in the luminosity layer).
And here we are...
That’s all folks!
Extra!
Neat trick to avoid Ringing at the Edges (from Chris [jcl6 / Reindeergraphics])
In
the final image above you can see that some of the texture has not been
removed at the top and bottom edges. Normally this is very difficult to avoid. If you put in a border of 50% grey, at least 100 pixels wide, all around the photo, and then do the FFT process the ringing will be contained to this border. At the end, just crop the border away and the ringing is gone.
Big Thankyou’s to:
Frannymae, Flora, Duv, Cameraken, jcl6 (reindeergraphics), jocker
Notes:
(*1)
You could avoid the problem with paper texture by taking a new
photograph of the image with the lighting set up to eliminate the
texture. If you do have this option, it is always a better solution
thantrying to “fix” the scan.
(*2)
The Alex Chirakov filter is free. It does introduce a small amount of
noise into the image however, as we are generally already dealing with
noisy images, this doesn’t seem to be too much of a problem. The
commercial FFT filters do not introduce noise at all but they (of course) are not free.
(*3) FFT download and installation.
- Access this page http://www.pages.drexel.edu/~avc25/archive.htm#FFT and download the “RGB Version of FFT/IFFT”.
-
You now have a file named “FFT_RGB_PlugIns……rar”. The “.rar” suffix
means that you need to unpack this with a program called WinRAR. If you
don’t already have WinRAR you can download it here http://www.rarlab.com/download.htm (official page)
-
With WinRAR, double-click “.rar” file and then open the “bin” folder.
There you will find two files (FFT_RGB.8bf and IFFT_RGB.8bf), extract
these to your Photoshop filters folder.
|
