'rolling-ball background substraction' in matlab

Is there a reliable 'rolling-ball background subtraction' scripts for matlab?

Answer 1

[Image Analyst]

Not exactly sure what that is, but morphological closing has been described as like a rolling ball. So to subtract that from the original image, the operation is called a bottom hat filter and is performed by the function imbothat() in the Image Processing Toolbox.

[Peyman Obeidy]

I tried both imbothat and imtophat, but I cant get only very dark spot (potentially separated ) in 'oldB16.jpg. The Bkg image is my background so the setting for the other image should peak no blak spot in bkg image.

[Peyman Obeidy]

Thank you Image Analyst.

After trying different background subtraction methods, I highly recommend this function. Very nice job. http://au.mathworks.com/matlabcentral/fileexchange/40266-sauvola-local-image-thresholding#feedbacks

Sauvola method. Happy to send the rest of code, let me know if you are interested.

[raym]

Rolling ball background was mentioned briefly in imerode:

https://www.mathworks.com/help/images/ref/imerode.html#bu7tf5c-1

It use erode to get a background of the raw image(better pre-process using gauss smooth to minimize the effect of very low pixels). The things matters is the shape of structure element and the offset. ImageJ's Rolling ball use a non-flat ball-shaped structure element. The optional paraboloid shape may correspond to the distorted ball-shaped structure element that has inharmonious height anf radius. An extreme end is very smoothed ball that is almost a flat pad, then it become the flat strel with no offset. The role of offset is to tell us that the a more distal-localized pixel should have less impact on the background value calculation, thus to minimize this effect, a additional offset is force added to this pixel just in case it has a very low value. [In matlab case, the origin is add the highest while the margin pixel add the lowest, thus all pixel are elevated.