Local Histogram Difference Method--A New Thresholding Method for Segmenting Noisy Gray-Scale Images--

Minoru ITO  Toshifumi WATANABE  

IEICE TRANSACTIONS (1976-1990)   Vol.E73   No.11   pp.1815-1817
Publication Date: 1990/11/25
Online ISSN: 
Print ISSN: 0000-0000
Type of Manuscript: Special Section LETTER (Special Issue on 1990 Autumn Natl. Conv. IEICE)
Category: Signals, Circuits and Images

Full Text: PDF>>
Buy this Article

This letter describes a new thresholding method for recognizing pattern regions in noisy gray-scale images. Many previously developed thresholding techniques occasionally lead to unreasonable results for noisy images with small pattern occupancies. In the proposed method, the threshold is evaluated on the basis of the difference function of a pair of local histograms for two windows set in the image. The zero-cross point of the function is independent of both the noise spectrum and the pattern ratio, and it coincides with the true threshold. This method is a powerful tool for thresholding noisy gray-scale images that have unknown pattern ratios.