[magick-users] RE: Again: On Installing & Configuring the FFT implementation for Noise removal

[Fred Weinhaus]

I may not be of much help, but you should get help from Sean Burke. 
However, be sure that you have properly installed the FFTW delegate 
library first:

For Mac installs, use sudo make install; otherwise make install



	Download the FFTW source:
	> curl -O http://www.fftw.org/fftw-3.1.2.tar.gz
	(That option is a capital "oh", not a zero.)
	Unzip and change into the source directory:
	> tar zxvf fftw-3.1.2.tar.gz && cd fftw-3.1.2
	Configure the build:
	> ./configure
	(Advanced users can call ./configure --help to display 
further enhancements.)
	Build FFTW and check the subsequent build:
	> make && make check
	Get a cup of coffee.  After the build completes, install it 
in the standard location:
	> sudo make install


	Download the most recent ImageMagick source:
	> curl -O ftp.imagemagick.org/pub/ImageMagick/ImageMagick.tar.gz
	Unzip and change directories:
	> tar zxvf ImageMagick.tar.gz && cd ImageMagick-6.3.7

	Configuring ImageMagick.  This step will require some thought 
on your part.   The default configuration is "Q16."   Q16 refers to 
how ImageMagick stores image data internally - in this case, image 
data is scaled from the original depth (usually 8 bit integers) to 16 
bit integers. This works well for 8 and 16 bit images, and not well 
for anything above 16 bits.  The average user will be more than happy 
with the 16 bit build.  In our lab, we use 32-bit tiffs for cell 
segmentation so we would need to use the "--with-quantum-depth=32" 
option.  An 8 bit (non-hdri - see next paragraph) build is not 
reccomended for Fourier work.  A "round trip" 8-bit image (the image 
produced by an inverse Fourier transform (IFT) of a Fourier transform 
(FT) of an image) on a Q16 build will have an error of up to 2%/pixel.

	> ./configure   (will be the default Q16 IM)
	Make the build.
	> make
	Get another cup of coffee.  Maybe two.
	Install ImageMagick:
	> sudo make install


Install imfft

	Change to the trunk directory of imfft:
	> cd imfft/trunk
	Build the test program, test the output:
	> make image && ./test_fft ../test_data/test_pattern6.png


As I recall, this might actually need to be
	> make demo && ./test_fft ../test_data/test_pattern6.png

Things have changed since this version. So the best advice is to 
contact Sean Burke if this does not work.

Fred