thresMinErrIter

[ XITE Reference Manual | XITE home ]

Name

thresMinErrIter - minimum error threshold, iteration method

Syntax

 #include <xite/threshold.h>

 int thresMinErrIter( histogram h, int tinit );

Description

thresMinErrIter calculates the threshold value by using the minimum error thresholding technique as in Kittler & Illingworth (2nd method) and Gonzalez & Wintz. Initially, the threshold value t is set according to tinit.

 0 <= tinit <= 255, then t = tinit
 tinit = -1, then t = the average of all the pixels
 tinit = -2, then t = the result from thresRidCal(h)

Iteratively, the new threshold is found by solving the equation (7.3-13) in Gonzalez & Wintz. The variances and the means are the variances and means above and belov the treshold. When the threshold value remains constant, iteration terminates.

The actual thresholding is not performed, see the routine "threshold".

Reference

 J. Kittler & J. Illingworth: "Minimum Error Thresholding"
Pattern Recognition, Vol 19, nr 1. 1986, pp. 41-47.
 Gonzales, Rafael C & Wintz, Paul:
 "Digital Image Processing"
 Addison-Wesley, 1977.
 Q.-Z. Ye & P.-E. Danielsson:
 "On Minimum Error Thresholding and Its Implementations"
 Pattern Recognition Letters 7,
 1988, pp. 201-206.

Return value

The threshold value. Pixel values up to, and including, the return value, is below the threshold.

Author

Olav Borgli, BLAB, ifi, UiO

Id

$Id: thresMinErr.c,v 1.25 1997/03/17 13:45:07 svein Exp $