Automatic threshold selection based on histogram modes and a discriminant criterion
Due to the unconstrained nature of image segmentation, the existing thresholding methods require considerable human intervention and pre-assumptions to determine appropriate threshold values. In this paper, a fully automatic thresholding method via histogram modal decomposition by data-dependent-systems methodology is presented. In this method, the histogram of an image is parametrically modeled by the power spectrum of an autoregressive model to provide vital information about histogram clusters. Utilizing the modal information, threshold values are then selected to maximize the between-class variance. The proposed method is validated by illustrative examples; comparison with the existing methods helps explain their differences and the superiority of the approach.
Machine Vision and Applications
Automatic threshold selection based on histogram modes and a discriminant criterion.
Machine Vision and Applications,
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p/4636