ComputerVision101_Week1_HistogramThresholding

university computervision week1 theory

Histogram Thresholding

Source:

• CV101_LectureNotes1b

Goal

Choose a threshold that separates foreground from background based on the histogram.

Manual approach

Inspect the histogram and choose a threshold when the distribution is bimodal.

IsoData idea

Define:

• m_L(t): mean of values <= t
• m_H(t): mean of values > t

Then update:

$$t=\frac{m_L(t)+m_H(t)}{2}$$

and iterate until the threshold stabilizes.