Title :
Efficient transformation for identifying global valley locations in 1D data
Author_Institution :
Machine Vision Group, Univ. of London, Egham
Abstract :
Described is a new transformation for finding global valleys in 1D distributions, with particular application to the thresholding of grey-scale images. The applied criterion function estimates the global significance of all the valleys that are located, and thus can cope with multimode distributions. The core transformation can be implemented in just two passes: hence the computational load is O(N) for an N-element distribution, which is optimal
Keywords :
geophysical signal processing; image classification; image segmentation; 1D data; N-element distribution; global valley locations; grey-scale images; multimode distributions;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20070120
