Can the rolling ball radius tool be used alone to reduce background in fluorescent images? Why or why not? If not, what are some additional methods to further remove background after rolling ball has been applied?
Yes, but this method applies the background subtraction locally. So, if the illumination is uneven, for example, it will correct for that. You do need to set the rolling ball radius to be at least the size of the largest object in the image. See below from the wiki article for this plugin.
This plugin tries to correct for uneven illuminated background by using a “rolling ball” algorithm.
A local background value is determined for every pixel by averaging over a very large ball around the pixel. This value is hereafter subtracted from the original image, hopefully removing large spatial variations of the background intensities. The radius should be set to at least the size of the largest object that is not part of the background.
This plugin implements (differently) the same algorithm as the one built-in ImageJ in the Process › Subtract background menu, but adds a useful Preview capability. Also, to display the background subtracted in a separate (new) window, hold the ALT key when pressing “OK” (Preview must be off).
The rolling-ball algorithm was inspired by Stanley Sternberg’s article, “Biomedical Image Processing”, IEEE Computer, January 1983.
I usually find this method the most accurate and unbiased to remove background. The theoretical explanation abovementioned by Christopher B O'Connell and the references provided are perfect.
I find Rolling-Ball advantageous because:
I) Minimizes user-intervention (potential sources of error and bias)
II) Since it's local, different values are subtracted to every pixel. Usually this is the best way to correct biomedical images unless you can assure a perfectly even illumination, which is often not the case (not due to the imaging system, but due to the sample)
Another option that some people use to remove background is to average several background ROIs (usually cell free regions, tissue free regions or un-labelled tissue) and perform an arithmetical subtraction of this value.
You can do this over the image in ImageJ by:
Process>Math>Subtract...
and input your value
Or, alternatively, after listing the values, remove this value to all your ROIs.
Again, I'd never advice this method since the user intervention could introduce errors (not perfect sampling of all the cell/tissue free regions; (un)intentional bias towards different experimental sets,...)
As a plus, in ImageJ you can find a plethora of different filters and de-noising tools (Under Process>Noise and Process>Filters, respectively). Although these aren't strictly speaking background removal tools, they can help to improve your images if used correctly. All of them are well explained here: