灰度直方图与阈值化算法详解-CSDN博客

Grayscale histogram and threshold

Histogram

Definition: Histogram is a pixel distribution function based on each gray level

Application:

set up the digitizing parameters

help to find an optimal threshold level (minimum of two peaks in the optimal for defining the boundary )

Binary Operation

fixed and adaptive threshold value

thresholding usually involves analyzing the histogram

an example of a threshold value is the mean intensity value

fixed threshold:

$g(x,y)=\left\{\begin{matrix} 0& f(x,y)< T\\ 1&f(x,y)\geq T\end{matrix}\right$

$g(x,y)=\left\{\begin{matrix}0& f(x,y)<T_{1}\\1& T_{1}\leq f(x,y)<T_{2}\\0& T_{2} \leq f(x,y)\end{matrix}\right$

Isodata Algorithm

reference：http://www.cnblogs.com/xiangshancuizhu/archive/2012/03/07/2384197.html

1. select and initial threshold $t_{0}$(mean intensity)

2. partition the image into two groups ( $R_{1}$ and $R_{2}$) using the $t_{0}$

3. calculate the mean intensity values $u_{1}$ and $u_{2}$ of the partition $R_{1}$ and $R_{2}$

4. select a new threshold : $t_{1}=(u_{1}+u_{2})/2$

5. repeat steps 2-4 until:$t_{1}=t_{n}$

OTSU algorithm

reference：http://blog.csdn.net/abcjennifer/article/details/6671288

http://www.cnblogs.com/skyseraph/archive/2010/12/21/1913058.html

selecting the lowest point between two classed(peaks)

Frequency (probability)

Mean

Total variance = standard deviation * standard deviation

其中$ \omega_0 \omega_1$分别为背景和前景像素数的概率，$ \mu_0 \mu_1$分别为背景和前景像素灰度均值

the criterion function involves between-classed variance to the total variance is defined as:

all possible thresholds are evaluated in this way, and the one that maximizes is $\eta$ chosen as the optimal threshold