Fuzzy multiscale edge detection (FMED) applied to automatic left ventricle boundary extraction

This paper describes a new fuzzy multiscale edge detection (FMED) technique. The wavelet transform is employed to produce the multiscale representation of the input signal. The process of developing appropriate multiscale fuzzy subsets describing the degree of 'edginess' for any pixel is explained. The new FMED technique provides an answer to the question "How to effectively combine the edge information in the different scales in order to end up with better edge detection results?" The paper compares the performance of FMED to the conventional gradient based Sobel edge detector and to Mallat's (1992) multiscale edge detection algorithm. The results show the superiority of FMED to these other methods.