Package qupath.opencv.tools

Class LocalNormalization

java.lang.Object

qupath.opencv.tools.LocalNormalization

public class LocalNormalization
extends Object

Methods to normalize the local image intensity within an image, to have (approximately) zero mean and unit variance. Calculations are made using Gaussian filters to give a smooth result.

Author:

Pete Bankhead

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	LocalNormalization.LocalNormalizationType	Helper class to store local normalization parameters.
static enum	LocalNormalization.NormalizationType	Local normalization type.
static class	LocalNormalization.SmoothingScale	2D or 3D Gaussian scale.

Constructor Summary

Constructors

Constructor	Description
LocalNormalization()

Method Summary

Modifier and Type	Method	Description
static void	gaussianNormalize(List<Mat> stack, LocalNormalization.LocalNormalizationType type, PixelCalibration cal, int border)	Apply local normalization to a stack of Mats representing a z-stack.
static void	gaussianNormalize2D(Mat mat, double sigma, double sigmaVariance, int border)	Apply local normalization to a 2D Mat.
static void	gaussianNormalize3D(List<Mat> stack, double sigmaX, double sigmaY, double sigmaZ, double varianceSigmaX, double varianceSigmaY, double varianceSigmaZ, int border)	Apply 3D normalization.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

LocalNormalization

public LocalNormalization()

Method Details

gaussianNormalize

public static void gaussianNormalize(List<Mat> stack,
 LocalNormalization.LocalNormalizationType type,
 PixelCalibration cal,
 int border)

Apply local normalization to a stack of Mats representing a z-stack.

Parameters:

stack -

type -

cal -

border -

gaussianNormalize2D

public static void gaussianNormalize2D(Mat mat,
 double sigma,
 double sigmaVariance,
 int border)

Apply local normalization to a 2D Mat.

Parameters:

mat -

sigma -

sigmaVariance -

border -

gaussianNormalize3D

public static void gaussianNormalize3D(List<Mat> stack,
 double sigmaX,
 double sigmaY,
 double sigmaZ,
 double varianceSigmaX,
 double varianceSigmaY,
 double varianceSigmaZ,
 int border)

Apply 3D normalization.

The algorithm works as follows:

1. A Gaussian filter is applied to a duplicate of the image
2. The filtered image is subtracted from the original
3. A local weighted variance estimate image is generated from the original image (by squaring, Gaussian filtering, subtracting the square of the smoothed image previously generated)
4. The square root of the weighted variance image is taken to give a normalization image, approximating a local standard deviation)
5. The subtracted image is divided by the value of the normalization image

The resulting image can be thought of as having a local mean of approximately zero and unit variance, although this is not exactly true; in practice there can be substantial differences. However, the approach aims to be simple, efficient and yield an image that does not introduce sharp discontinuities by is reliance on Gaussian filters.

Parameters:

stack - image z-stack, in which each element is a 2D (x,y) slice

sigmaX - horizontal Gaussian filter sigma

sigmaY - vertical Gaussian filter sigma

sigmaZ - z-dimension Gaussian filter sigma

varianceSigmaX - horizontal Gaussian filter sigma for variance estimation

varianceSigmaY - vertical Gaussian filter sigma for variance estimation

varianceSigmaZ - z-dimension Gaussian filter sigma for variance estimation

border - border padding method to use (see OpenCV for definitions)