boofcv.alg.filter.binary

Class GThresholdImageOps

java.lang.Object

boofcv.alg.filter.binary.GThresholdImageOps

public class GThresholdImageOps
extends java.lang.Object

Weakly typed version of ThresholdImageOps.

Constructor Summary

Constructors

Constructor and Description
GThresholdImageOps()

Method Summary

Modifier and Type	Method and Description
static int	computeEntropy(ImageGray input, int minValue, int maxValue) Computes a threshold which maximizes the entropy between the foreground and background regions.
static int	computeEntropy(int[] histogram, int length, int totalPixels) Computes a threshold which maximizes the entropy between the foreground and background regions.
static int	computeOtsu(ImageGray input, int minValue, int maxValue) Computes the variance based threshold using Otsu's method from an input image.
static int	computeOtsu(int[] histogram, int length, int totalPixels) Computes the variance based Otsu threshold from a histogram directly.
static <T extends ImageGray> GrayU8	localBlockMinMax(T input, GrayU8 output, int radius, double scale, boolean down, double textureThreshold) Applies a threshold to an image by computing the min and max values in a regular grid across the input image.
static <T extends ImageGray> GrayU8	localGaussian(T input, GrayU8 output, int radius, double scale, boolean down, T work1, ImageGray work2) Thresholds the image using a locally adaptive threshold that is computed using a local square region centered on each pixel.
static <T extends ImageGray> GrayU8	localSauvola(T input, GrayU8 output, int radius, float k, boolean down) Applies Sauvola thresholding to the input image.
static <T extends ImageGray> GrayU8	localSquare(T input, GrayU8 output, int radius, double scale, boolean down, T work1, T work2) Thresholds the image using a locally adaptive threshold that is computed using a local square region centered on each pixel.
static <T extends ImageGray> GrayU8	threshold(T input, GrayU8 output, double threshold, boolean down) Applies a global threshold across the whole image.

Methods inherited from class java.lang.Object

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

Constructor Detail

GThresholdImageOps

public GThresholdImageOps()

Method Detail

computeOtsu

public static int computeOtsu(ImageGray input,
                              int minValue,
                              int maxValue)

Computes the variance based threshold using Otsu's method from an input image. Internally it uses computeOtsu(int[], int, int) and GImageStatistics.histogram(ImageGray, int, int[])

Parameters:

input - Input gray-scale image

minValue - The minimum value of a pixel in the image. (inclusive)

maxValue - The maximum value of a pixel in the image. (inclusive)

Returns:

Selected threshold.

computeOtsu

public static int computeOtsu(int[] histogram,
                              int length,
                              int totalPixels)

Computes the variance based Otsu threshold from a histogram directly. The threshold is selected by minimizing the spread of both foreground and background pixel values.

Parameters:

histogram - Histogram of pixel intensities.

length - Number of elements in the histogram.

totalPixels - Total pixels in the image

Returns:

Selected threshold

computeEntropy

public static int computeEntropy(ImageGray input,
                                 int minValue,
                                 int maxValue)

Computes a threshold which maximizes the entropy between the foreground and background regions. See computeEntropy(int[], int, int) for more details.

Parameters:

input - Input gray-scale image

minValue - The minimum value of a pixel in the image. (inclusive)

maxValue - The maximum value of a pixel in the image. (inclusive)

Returns:

Selected threshold.

See Also:

GImageStatistics.histogram(ImageGray, int, int[])

computeEntropy

public static int computeEntropy(int[] histogram,
                                 int length,
                                 int totalPixels)

Computes a threshold which maximizes the entropy between the foreground and background regions. See [1] for algorithmic details, which cites [2].

[1] E.R. Davies "Machine Vision Theory Algorithms Practicalities" 3rd Ed. 2005. pg. 124
[2] Hannah, Ian, Devesh Patel, and Roy Davies. "The use of variance and entropic thresholding methods for image segmentation." Pattern Recognition 28.8 (1995): 1135-1143.

Parameters:

histogram - Histogram of pixel intensities.

length - Number of elements in the histogram.

totalPixels - Total pixels in the image

Returns:

Selected threshold

threshold

public static <T extends ImageGray> GrayU8 threshold(T input,
                                                     GrayU8 output,
                                                     double threshold,
                                                     boolean down)

Applies a global threshold across the whole image. If 'down' is true, then pixels with values ≤ to 'threshold' are set to 1 and the others set to 0. If 'down' is false, then pixels with values > to 'threshold' are set to 1 and the others set to 0.

Parameters:

input - Input image. Not modified.

output - (Optional) Binary output image. If null a new image will be declared. Modified.

threshold - threshold value.

down - If true then the inequality ≤ is used, otherwise if false then > is used.

Returns:

binary image.

localSquare

public static <T extends ImageGray> GrayU8 localSquare(T input,
                                                       GrayU8 output,
                                                       int radius,
                                                       double scale,
                                                       boolean down,
                                                       T work1,
                                                       T work2)

Thresholds the image using a locally adaptive threshold that is computed using a local square region centered on each pixel. The threshold is equal to the average value of the surrounding pixels times the scale. If down is true then b(x,y) = I(x,y) ≤ T(x,y) * scale ? 1 : 0. Otherwise b(x,y) = I(x,y) > T(x,y) * scale ? 0 : 1

NOTE: Internally, images are declared to store intermediate results. If more control is needed over memory call the type specific function.

Parameters:

input - Input image.

output - (optional) Output binary image. If null it will be declared internally.

radius - Radius of square region.

scale - Scale factor used to adjust threshold

down - Should it threshold up or down.

work1 - (Optional) Internal workspace. Can be null

work2 - (Optional) Internal workspace. Can be null

Returns:

binary image.

localGaussian

public static <T extends ImageGray> GrayU8 localGaussian(T input,
                                                         GrayU8 output,
                                                         int radius,
                                                         double scale,
                                                         boolean down,
                                                         T work1,
                                                         ImageGray work2)

Thresholds the image using a locally adaptive threshold that is computed using a local square region centered on each pixel. The threshold is equal to the gaussian weighted sum of the surrounding pixels times the scale. If down is true then b(x,y) = I(x,y) ≤ T(x,y) * scale ? 1 : 0. Otherwise b(x,y) = I(x,y) > T(x,y) * scale ? 0 : 1

NOTE: Internally, images are declared to store intermediate results. If more control is needed over memory call the type specific function.

Parameters:

input - Input image.

output - (optional) Output binary image. If null it will be declared internally.

radius - Radius of square region.

scale - Scale factor used to adjust threshold

down - Should it threshold up or down.

work1 - (Optional) Internal workspace. Can be null

work2 - (Optional) Internal workspace. Can be null

Returns:

binary image.

localSauvola

public static <T extends ImageGray> GrayU8 localSauvola(T input,
                                                        GrayU8 output,
                                                        int radius,
                                                        float k,
                                                        boolean down)

Applies Sauvola thresholding to the input image. Intended for use with text image.

Parameters:

input - Input image.

output - (optional) Output binary image. If null it will be declared internally.

radius - Radius of local region. Try 15

k - Positive parameter used to tune threshold. Try 0.3

down - Should it threshold up or down.

Returns:

binary image

See Also:

ThresholdSauvola

localBlockMinMax

public static <T extends ImageGray> GrayU8 localBlockMinMax(T input,
                                                            GrayU8 output,
                                                            int radius,
                                                            double scale,
                                                            boolean down,
                                                            double textureThreshold)

Applies a threshold to an image by computing the min and max values in a regular grid across the input image. See ThresholdSquareBlockMinMax for the details.

Parameters:

input - Input image.

output - (optional) Output binary image. If null it will be declared internally.

radius - Radius of square region.

scale - Scale factor used to adjust threshold

down - Should it threshold up or down.

textureThreshold - If the min and max values are within this threshold the pixel will be set to 1.

Returns:

Binary image