I have done cattle image segmentation and edge detection of the cattle image attached here. I have equally obtained the boundary pixel along x and y axis where the pixel value equal to 1. I want to do shape recognition and description of the cattle image attached here. I need an explanation on the next step to take and the MATLAB code for the description and representation using 1-D Fourier descriptor and bounding box.
The bounding box is trivial. In one of your other threads, an (x,y) pixel-index array of the contour has been suggested. Using min() and max() on that array will give you the bounding box.
Converting to a 1D signal... Chain-coding is a possibility, but better is to imagine yourself 'driving' along the contour. Straight ahead is 0; a 45 degree turn is -1 or +1; a 90 degree turn is -2 or +2; a 135 degree turn is -3 or +3.
That should give you a 1D signal, and if you coded it right, you end at the pixel you started with. Going to the Fourier domain, does not make much sense to me, but could be done of course.
B.t.w. if you zip or rar that 1D signal, you get a highly efficient shape descriptor!
thank you Mr. Lambert Zijp. Please can i get any MATLAB code from you that i use for the shape recognition.
http://www.mathworks.co.uk/matlabcentral/fileexchange/12746-elliptical-fourier-shape-descriptors
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing