How to represent the skeleton of a mask as an equation of a line?

1 view (last 30 days)
My masks are assumed to always take the following shape, and I need to represent them with minimum information (x_start, x_end, and an equation for all points in between). The only variance in the shape of these masks are that they can be 2-4 times thicker, and the curvature at the top can be more pronounced (top). For "short" masks - where the length of the object is shorter - I can simply use 'fitlm'. For "longer" masks, fitting a fourier series to the mask's skeleton works reasonably well. For masks like this, though, I can't get an accurate representation (bottom). I'm thinking that polar coordinates would be an easy fix?
How can I get a more thorough representation of this object? Here is the code for fitting a fourier series.
[y, x] = find( bwmorph( BW, 'thin', Inf ) )
B = horzcat( x, y );
fittedFunc = fit( B( :, 1 ), B( :, 2 ), 'fourier1' ); % Only gets worse with higher degrees.
% Create a function handle from the fit.
fx = strcat( '@(x)', formula( fittedFunc ) );
names = coeffnames( fittedFunc );
values = coeffvalues( fittedFunc );
for idx = 1:length( names )
fx = strrep( fx, names( idx ), num2str( values( idx ) ) );
end
fs = '';
for idx = 1:length( fx{ 1 } )
if ~isstrprop( fx{ 1 }( idx ), 'wspace' )
fs = strcat( fs, fx{ 1 }( idx ) );
end
end
end
eq = str2func( fs );
% Testing
[~, xBW] = find( BW );
xplot = linspace( min( xBW ), max( xBW ), 1000 );
yplot = eq( xplot );
figure; imshow( BW ); hold on; plot( xplot, yplot, 'r.-' );
  2 Comments
Image Analyst
Image Analyst on 4 Feb 2020
"I need to represent them with minimum information" WHy??? What's wrong with the binary image matrix? Either in memory, or on disk as a PNG image file, it does not take up much memory. Why do you need to complicate things by converting to a different format?
Dominik Mattioli
Dominik Mattioli on 4 Feb 2020
Edited: Dominik Mattioli on 4 Feb 2020
Primarily because there are context-dependent properties of the object that I use for a more complex analysis involving other objects, and ultimately those properties and analysis are derived from the minimum information of those objects as they relate to one other. I can go into great detail, if you'd like, but I figured it was not relevant to the question.

Sign in to comment.

Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!