This function takes as parameters a grayscale image (real valued matrix of size x*w*1) and two thresholds (low and high), and returns the hysteresis thresholded version.
Hysteresis thresholding performs the following:
- every value below tLo is set to 0
- every value above tHi is set to 1
- the rest of the pixels are set to 1 if they are 4-way connected to any other 1-valued blob (area), 0 otherwise
It should be pretty straightforward to implement an 8-way connection check, if you want.
To compile and use perform the following:
I wasn't able to get this to work myself, but programing the functionality in Matlab is very easy
highmask = image>tHi;
lowmask = bwlabel(~(image<tLo));
final = ismember(lowmask,unique(lowmask(highmask)));
There seems to be only 1 set of forward-backward propagation set. It'd be nice if the code was extended to do recursively until nothing changes anymore as suggested on http://users.ecs.soton.ac.uk/msn/book/new_demo/thresholding/
I guess its really just putting the propagation part in a while loop with checking if any pixel switched from 0 to 1.
Great function otherwise!
make sure grayscale image is of type double
Hi when I compiled and executed this code, I got the following error. I invoked it as TI = hysthresh(GI,0.1,0.4), where GI is a grayscale image.
??? Error using ==> hysthresh
YPRIME requires that Y be a 4 x 1 vector.
I also tried commenting this part in the code and recompiling the code. However, it ran out of memory.
Corrected a type in the title
Inspired: Circular Matrix computation