Thresholding methods considering the quantization error of an image
Iwao Sekita, Takio Kurita, Nobuyuki Otsu, Nabih N. Abdelmalek · Systems and Computers in Japan · 1996
Abstract Thresholding techniques are fundamental for region segmentation of a gray level image. It is often realistic to assume that gray pixel levels are subject to a mixture of normal distributions. This paper points out problems caused by the underestimation of variances from a frequency distribution and solves the problems by transforming a frequency distribution into a piecewise‐continuous distribution based on the maximum entropy criterion. This transformation improves conventional thresholding techniques. Experimental results showed the effectiveness of the transformation.