Thank you for your reply.
I think the following link will be useful:
(an optimization of this algorithm).
In image processing, expecially in fingerprint recognition, the
result of the cited algorithm has to be "refined" to find minutiae
because the algorithm can detect "false" minutiae:
Steve Eddins wrote:
> Luigi <luigi.rosa@(at)tiscali.it> writes:
>> What is the algorithm used to perform the binary image
> For thinning, bwmorph divides the image pixels into two subfields
> in a
> checkerboard fashion. Each iteration has two subiterations. In
> the first
> subiteration, pixels in the first subfield are deleted if and only
> if they
> satisfy three conditions designed to eliminate pixels while
> topology. In the second subiteration, pixels in the second
> subfield are
> deleted if and only if they satisfy a slightly different set of
> conditions. The conditions are tested using the applylut function
> precomputed lookup tables. (Using lookup tables is efficient, but
> it makes
> it hard to understand the algorithm just by looking at the code
> bwmorph.) If the user specifies an infinite number of iterations,
> iterations are repeated until the image stops changing.
> The conditions are a bit hard to write in plain text. I'll send
> you a Word
> file with the equations.
> Reference: Louisa Lam, Seong-Whan Lee, and Ching Y. Suen, "Thinning
> Methodologies -- A Comprehensive Survey," IEEE Transactions on
> Analysis and Machine Intelligence, vol. 14, no. 9, September 1992,
> p. 879,
> bottom of first column through top of second column.
> I'll have these algorithm notes added to the bwmorph reference
> Steve Eddins
> Development Manager, Image Processing Group
> The MathWorks, Inc.