Page 322 - Fundamentals of Light Microscopy and Electronic Imaging

SIGNAL-TO-NOISE RATIO 305

where
C object counts in ADU where C object C total C background
o
C object noise in ADU
o
n number of pixels in measured object area
p number of pixels in measured background area
2
2
variance of background pixels as (SD) in ADU
g gain (electrons/ADU)
The C and 2 (background) measurements are taken from selected object and adjacent
o
background ROIs using the Statistics or Measure ROI command in the analysis software.
The units are in ADU with the correction for electrons appearing as gain in the denomi-
nator. The background variance is the calculated statistical variance of the pixel sample
(the square of the standard deviation) and is obtained from the Measure ROI command;
it is not the Poissonian mean that was discussed earlier under photon shot noise. For con-
venience, set up these equations and related tables of variables in a spreadsheet.

Exercise: Flat-Field Correction
and Determination of S/N Ratio

In this exercise you are to correct a raw image by flat fielding, extract numeric
values from a structure in the image, and determine the signal-to-noise ratio. You
will need access to Adobe Photoshop software on a computer. Although this
manipulation is somewhat awkward in Photoshop, the exercise will make you
think about what you are doing and reinforce the processing operations. Prepare
raw, flat and dark images with a CCD camera, convert them to 8 bit data type
(byte), and save them in TIFF format on the hard drive. In some cases, a supervi-
sor may choose to do this for you. A flat-field correction is obtained by perform-
ing the following steps:

Corrected image M [(R D)/(F D)]

a. Subtract the dark frame from the raw frame (R D).
In PS: Image/Calculations: Source 1 raw; source 2 dark; blending
difference; opacity 100%; result new; channel new; save the
new image.
b. Subtract the dark frame from the flat frame (F D).
In PS: same as above and save the image.
c. Determine the average pixel value M in the corrected flat frame (F D).
In PS: Image/Histogram: Record the mean value from data table.
d. Divide the corrected raw image frame by the corrected flat-field frame
(R D)/(F D).
In PS:You cannot divide images in PS, so take the inverse of the corrected
flat frame and multiply it by the corrected raw frame. Use the menus: