## TF DNS - Transfer Function - Delete noise background

(11/5/15)

PURPOSE

Calculate the noise background from a rotationally averaged profile of power spectrum and correct it. The background noise is assumed to have a Gaussian profile: a1 * exp[ - (k/a2)**2]+a3. The output file is used by 'TF DEV' for envelope function fitting. Further info on CTF related operations in SPIDER.   Example.

 TF DEV [Transfer Function - Determine Envelope function] TF C [Transfer Function - Generate a straight, complex, CTF correction image] TF DDF [Transfer Function - Determine Defocus & amplitude contrast] TF LM4 [Transfer Function - Determine CTF envelope B-factor and noise parameters] TF DNS [Transfer Function - Delete noise background] CTF FIND [Contrast Transfer Function - Estimation of CTF parameters]

USAGE

.OPERATION: TF DNS [minima]
[Denoise one dimensional power spectrum. This operation can return one optional register variable which contains the Number of minima found.

.1D POWER SPECTRUM IMAGE FILE: ro008
[Enter name of a file containing 1D half-profile of power spectrum.]

.MAX SPATIAL FREQUENCY [1/A]: 0.104
[Enter the spatial frequency limit in units of 1 / Angstroms. The maximum spatial frequency is 1/(2*pixelsize), where pixelsize is the size of the pixel in Angstroms.]

.SEARCH NEIGHBORHOOD DISTANCE: 5
[Enter the local distance for defining minima. This is half the width that must be above the current minimum.]

At this point, minima are located and printed out. The first column contains the keys, the second column is the location of each minimum (value is interpolated between pixels), the third column has the same location in spatial frequency units. The last column gives the amplitude of the minimum. E.g.:
``` CURVE HAS: 8 MINIMA:
(PIXELS) (FREQ = 1/A)
1 42.00 0.0164 0.0732
2 175.84 0.0687 0.0300
3 249.61 0.0975 0.0251
4 303.99 0.1187 0.3888
5 349.06 0.1364 0.2255
6 395.03 0.1543 0.0160
7 430.32 0.1681 0.4989
8 467.99 0.1828 0.1609
```

.CHANGE SEARCH NEIGHBORHOOD? (Y/N): No
[ If "Y", the above operation will be repeated. Smaller search areas may identify noise as minima, resulting in spurious minima. Large search areas can smooth out and overlook actual minima if they are too small. Increase the search neighborhood if your data is very noisy; decrease it if you have many small minima. Normally the number of minima can be judged by eye. So, the neighborhood distance can be changed to get the number of minima that you expect.]

.LIST OF MINIMA USED TO DEFINE NOISE CURVE: 2-8
[Enter the numbers of minima that you wish to use in defining noise curve.]

If only one minimum is included in the calculation, SPIDER asks:

.A2 VALUE [1/A]: 0.05
[Enter the halfwidth of the Gaussian noise profile. This value is similar for all the micrographs recorded under the same conditions. Thus, it can be guessed from other calculations with more than minima. Or, it can be guestimated by checking the output file.]

.DENOISED PROFILE FILE: rod008
[Enter name of file to store the 1D background noise corrected profile.]

NOTES

1. The background noise is assumed to have a Gaussian profile. The background-corrected 1D profile has all the minima brought down to zero without changing the positions of the minima.

SUBROUTINES

NOISE, DEFO003

CALLER

UTIL1