Binarize adaptively

Transforms any gray-level or color image to a binary image using one of the selected adaptive thresholding algorithms.

Inputs

• image: Image to be thresholded. Color images will be converted to gray scale before applying the threshold.

• windowSize: Size of the local window for adaptive threshold calculation.

• absoluteLevel: Absolute threshold. The effect of this value depends on algorithm.

• relativeLevel: Relative threshold. The effect of this value depends on algorithm.

• algorithm: Adaptive thresholding algorithm.

• invert: If true, inverts the result so that all foreground pixels become background and vice versa.

Outputs

• image: Binary image

enum Algorithm

Thresholding algorithms.

Values:

enumerator RelativeToMean

Local threshold is calculated by multiplying the mean gray level around a pixel by relativeLevel and adding absoluteThreshold to the result.

enumerator RelativeToMeanAndStd

Local threshold is calculated by adding relativeThreshold times the standard deviation of the local gray levels to the mean gray level.

absoluteLevel will be added to the result.

enumerator WeightedMeanAndVariance

Local threshold t is calculated as \( t = \mu (1 + r (\sigma/\sigma_{\mathrm max} - 1)) + a \), where r stands for relativeThreshold, a for absoluteThreshold, and \( \sigma_{\mathrm max} \) for the maximum possible standard deviation, which is assumed to be 128.

In document image binarization, a good value for relativeThreshold is 0.34, whereas absoluteThreshold is typically zero.