Image Processing Tools → Image Enhancement

• BalanceContrastEnhancement
• CorrectVignetting
• DirectDecorrelationStretch
• GammaCorrection
• GaussianContrastStretch
• HistogramEqualization
• HistogramMatching
• HistogramMatchingTwoImages
• MinMaxContrastStretch
• PanchromaticSharpening
• PercentageContrastStretch
• SigmoidalContrastStretch
• StandardDeviationContrastStretch

BalanceContrastEnhancement

This tool can be used to reduce colour bias in a colour composite image based on the technique described by Liu (1991). Colour bias is a common phenomena with colour images derived from multispectral imagery, whereby a higher average brightness value in one band results in over-representation of that band in the colour composite. The tool essentially applies a parabolic stretch to each of the three bands in a user specified RGB colour composite, forcing the histograms of each band to have the same minimum, maximum, and average values while maintaining their overall histogram shape. For greater detail on the operation of the tool, please see Liu (1991). Aside from the names of the input and output colour composite images, the user must also set the value of E, the desired output band mean, where 20 < E < 235.

Reference:

Liu, J.G. (1991) Balance contrast enhancement technique and its application in image colour composition. International Journal of Remote Sensing, 12:10.

See Also: DirectDecorrelationStretch, HistogramMatching, HistogramMatchingTwoImages, HistogramEqualization, GaussianContrastStretch

Parameters:

Python function:

wbt.balance_contrast_enhancement(
    i, 
    output, 
    band_mean=100.0, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=BalanceContrastEnhancement -v ^
--wd="/path/to/data/" --input=image.tif -o=output.tif ^
--band_mean=120 

Source code on GitHub

Author: Dr. John Lindsay

Created: 19/07/2017

Last Modified: 30/01/2020

CorrectVignetting

This tool can be used to reduce vignetting within an image. Vignetting refers to the reducuction of image brightness away from the image centre (i.e. the principal point). Vignetting is a radiometric distortion resulting from lens characteristics. The algorithm calculates the brightness value in the output image (BVout) as:

BVout = BVin / [cos^n(arctan(d / f))]

Where d is the photo-distance from the principal point in millimetres, f is the focal length of the camera, in millimeters, and n is a user-specified parameter. Pixel distances are converted to photo-distances (in millimetres) using the specified image width, i.e. distance between left and right edges (mm). For many cameras, 4.0 is an appropriate value of the n parameter. A second pass of the image is used to rescale the output image so that it possesses the same minimum and maximum values as the input image.

If an RGB image is input, the analysis will be performed on the intensity component of the HSI transform.

Parameters:

Python function:

wbt.correct_vignetting(
    i, 
    pp, 
    output, 
    focal_length=304.8, 
    image_width=228.6, 
    n=4.0, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=CorrectVignetting -v ^
--wd="/path/to/data/" -i=input.tif --pp=princ_pt.shp ^
-o=output.tif --focal_length=304.8 --image_width=228.6 ^
-n=4.0 

Source code on GitHub

Author: Dr. John Lindsay

Created: 24/04/2018

Last Modified: 22/10/2019

DirectDecorrelationStretch

The Direct Decorrelation Stretch (DDS) is a simple type of saturation stretch. The stretch is applied to a colour composite image and is used to improve the saturation, or colourfulness, of the image. The DDS operates by reducing the achromatic (grey) component of a pixel's colour by a scale factor (k), such that the red (r), green (g), and blue (b) components of the output colour are defined as:

r_k = r - k min(r, g, b)

g_k = g - k min(r, g, b)

b_k = b - k min(r, g, b)

The achromatic factor (k) can range between 0 (no effect) and 1 (full saturation stretch), although typical values range from 0.3 to 0.7. A linear stretch is used afterwards to adjust overall image brightness. Liu and Moore (1996) recommend applying a colour balance stretch, such as BalanceContrastEnhancement before using the DDS.

Reference:

Liu, J.G., and Moore, J. (1996) Direct decorrelation stretch technique for RGB colour composition. International Journal of Remote Sensing, 17:5, 1005-1018.

See Also: CreateColourComposite, BalanceContrastEnhancement

Parameters:

Python function:

wbt.direct_decorrelation_stretch(
    i, 
    output, 
    k=0.5, 
    clip=1.0, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=DirectDecorrelationStretch -v ^
--wd="/path/to/data/" --input=image.tif -o=output.tif -k=0.4 

Source code on GitHub

Author: Dr. John Lindsay

Created: 21/07/2017

Last Modified: 30/01/2020

GammaCorrection

This tool performs a gamma colour correction transform on an input image (--input), such that each input pixel value (z_in^{) is mapped to the corresponding output value (z_out) as:}

z_out = z_in^gamma

The user must specify the value of the gamma parameter. The input image may be of either a greyscale or RGB colour composite data type.

Parameters:

Python function:

wbt.gamma_correction(
    i, 
    output, 
    gamma=0.5, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=GammaCorrection -v ^
--wd="/path/to/data/" -i=input.tif -o=output.tif --gamma=0.5 

Source code on GitHub

Author: Dr. John Lindsay

Created: 13/07/2017

Last Modified: 22/10/2019

GaussianContrastStretch

This tool performs a Gaussian stretch on a raster image. The observed histogram of the input image is fitted to a Gaussian histogram, i.e. normal distribution. A histogram matching technique is used to map the values from the input image onto the output Gaussian distribution. The user must the number of tones (--num_tones) used in the output image.

This tool is related to the more general HistogramMatching tool, which can be used to fit any frequency distribution to an input image, and other contrast enhancement tools such as HistogramEqualization, MinMaxContrastStretch, PercentageContrastStretch, SigmoidalContrastStretch, and StandardDeviationContrastStretch.

See Also: HistogramEqualization, MinMaxContrastStretch, PercentageContrastStretch, SigmoidalContrastStretch, StandardDeviationContrastStretch, HistogramMatching

Parameters:

Python function:

wbt.gaussian_contrast_stretch(
    i, 
    output, 
    num_tones=256, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=GaussianContrastStretch -v ^
--wd="/path/to/data/" -i=input.tif -o=output.tif ^
--num_tones=1024 

Source code on GitHub

Author: Dr. John Lindsay

Created: 21/05/2018

Last Modified: 22/10/2019

HistogramEqualization

This tool alters the cumulative distribution function (CDF) of a raster image to match, as closely as possible, the CDF of a uniform distribution. Histogram equalization works by first calculating the histogram of the input image. This input histogram is then converted into a CDF. Each grid cell value in the input image is then mapped to the corresponding value in the uniform distribution's CDF that has an equivalent (or as close as possible) cumulative probability value. Histogram equalization provides a very effective means of performing image contrast adjustment in an efficient manner with little need for human input.

The user must specify the name of the input image to perform histogram equalization on. The user must also specify the number of tones, corresponding to the number of histogram bins used in the analysis.

HistogramEqualization is related to the HistogramMatchingTwoImages tool (used when an image's CDF is to be matched to a reference CDF derived from a reference image). Similarly, HistogramMatching, and GaussianContrastStretch are similarly related tools frequently used for image contrast adjustment, where the reference CDFs are uniform and Gaussian (normal) respectively.

Notes:

• The algorithm can introduces gaps in the histograms (steps in the CDF). This is to be expected because the histogram is being distorted. This is more prevalent for integer-level images.
• Histogram equalization is not appropriate for images containing categorical (class) data.

See Also: HistogramMatching, HistogramMatchingTwoImages, GaussianContrastStretch

Parameters:

Python function:

wbt.histogram_equalization(
    i, 
    output, 
    num_tones=256, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=HistogramEqualization -v ^
--wd="/path/to/data/" -i=input.tif -o=output.tif ^
--num_tones=1024 

Source code on GitHub

Author: Dr. John Lindsay

Created: 26/08/2017

Last Modified: 30/01/2020

HistogramMatching

This tool alters the cumulative distribution function (CDF) of a raster image to match, as closely as possible, the CDF of a reference histogram. Histogram matching works by first calculating the histogram of the input image. This input histogram and reference histograms are each then converted into CDFs. Each grid cell value in the input image is then mapped to the corresponding value in the reference CDF that has an equivalent (or as close as possible) cumulative probability value. Histogram matching provides the most flexible means of performing image contrast adjustment.

The reference histogram must be specified to the tool in the form of a text file (.txt), provided using the --histo_file flag. This file must contain two columns (delimited by a tab, space, comma, colon, or semicolon) where the first column contains the x value (i.e. the values that will be assigned to the grid cells in the output image) and the second column contains the frequency or probability. Note that 1) the file must not contain a header row, 2) each x value/frequency pair must be on a separate row, and 3) the frequency/probability must not be cumulative (i.e. the file must contain the histogram and not the CDF). The CDF will be computed for the reference histogram automatically by the tool. It is possible to create this type of histogram using the wide range of distribution tools available in most spreadsheet programs (e.g. Excel or LibreOffice's Calc program). You must save the file as a text-only (ASCII) file.

HistogramMatching is related to the HistogramMatchingTwoImages tool, which can be used when a reference CDF can be derived from a reference image. HistogramEqualization and GaussianContrastStretch are similarly related tools frequently used for image contrast adjustment, where the reference CDFs are uniform and Gaussian (normal) respectively.

Notes:

• The algorithm can introduces gaps in the histograms (steps in the CDF). This is to be expected because the histogram is being distorted. This is more prevalent for integer-level images.
• Histogram matching is not appropriate for images containing categorical (class) data.
• This tool is not intended for images containing RGB data. If this is the case, the colour channels should be split using the SplitColourComposite tool.

See Also: HistogramMatchingTwoImages, HistogramEqualization, GaussianContrastStretch, SplitColourComposite

Parameters:

Python function:

wbt.histogram_matching(
    i, 
    histo_file, 
    output, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=HistogramMatching -v ^
--wd="/path/to/data/" -i=input1.tif --histo_file=histo.txt ^
-o=output.tif 

Source code on GitHub

Author: Dr. John Lindsay

Created: 14/09/2017

Last Modified: 13/10/2018

HistogramMatchingTwoImages

This tool alters the cumulative distribution function (CDF) of a raster image to match, as closely as possible, the CDF of a reference image. Histogram matching works by first calculating the histograms of the input image (i.e. the image to be adjusted) and the reference image. These histograms are then converted into CDFs. Each grid cell value in the input image is then mapped to the corresponding value in the reference CDF that has the an equivalent (or as close as possible) cumulative probability value. A common application of this is to match the images from two sensors with slightly different responses, or images from the same sensor, but the sensor's response is known to change over time.The size of the two images (rows and columns) do not need to be the same, nor do they need to be geographically overlapping.

HistogramMatchingTwoImages is related to the HistogramMatching tool, which can be used when a reference CDF is used directly rather than deriving it from a reference image. HistogramEqualization and GaussianContrastStretch are similarly related tools, where the reference CDFs are uniform and Gaussian (normal) respectively.

The algorithm may introduces gaps in the histograms (steps in the CDF). This is to be expected because the histograms are being distorted. This is more prevalent for integer-level images. Histogram matching is not appropriate for images containing categorical (class) data. It is also not intended for images containing RGB data, in which case, the colour channels should be split using the SplitColourComposite tool.

See Also: HistogramMatching, HistogramEqualization, GaussianContrastStretch, SplitColourComposite

Parameters:

Python function:

wbt.histogram_matching_two_images(
    input1, 
    input2, 
    output, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=HistogramMatchingTwoImages -v ^
--wd="/path/to/data/" --i1=input1.tif --i2=input2.tif ^
-o=output.tif 

Source code on GitHub

Author: Dr. John Lindsay

Created: 31/08/2017

Last Modified: 13/10/2018

MinMaxContrastStretch

This tool performs a minimum-maximum contrast stretch on a raster image. This operation maps each grid cell value in the input raster image (z) onto a new scale that ranges from the user-specified lower-tail clip value (min_val) to the upper-tail clip value (max_val), with the specified number of tonal values (num_tones), such that:

z_out = ((z_in – min_val)/(max_val – min_val)) x num_tones

where z_out is the output value. Notice that any values in the input image that are less than min_val are assigned a value of min_val in the output image. Similarly, any input values greater than max_val are assigned a value of max_val in the output image.

This is a type of linear contrast stretch with saturation at the tails of the frequency distribution. This is the same kind of stretch that is used to display raster type data on the fly in many GIS software packages, such that the lower and upper tail values are set using the minimum and maximum display values and the number of tonal values is determined by the number of palette entries.

See Also: GaussianContrastStretch, HistogramEqualization, PercentageContrastStretch, SigmoidalContrastStretch, StandardDeviationContrastStretch

Parameters:

Python function:

wbt.min_max_contrast_stretch(
    i, 
    output, 
    min_val, 
    max_val, 
    num_tones=256, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=MinMaxContrastStretch -v ^
--wd="/path/to/data/" -i=input.tif -o=output.tif ^
--min_val=45.0 --max_val=200.0 --num_tones=1024 

Source code on GitHub

Author: Dr. John Lindsay

Created: 13/07/2017

Last Modified: 30/01/2020

PanchromaticSharpening

Panchromatic sharpening, or simply pan-sharpening, refers to a range of techniques that can be used to merge finer spatial resolution panchromatic images with coarser spatial resolution multi-spectral images. The multi-spectral data provides colour information while the panchromatic image provides improved spatial information. This procedure is sometimes called image fusion. Jensen (2015) describes panchromatic sharpening in detail.

Whitebox provides two common methods for panchromatic sharpening including the Brovey transformation and the Intensity-Hue-Saturation (IHS) methods. Both of these techniques provide the best results when the range of wavelengths detected by the panchromatic image overlap significantly with the wavelength range covered by the three multi-spectral bands that are used. When this is not the case, the resulting colour composite will likely have colour properties that are dissimilar to the colour composite generated by the original multispectral images. For Landsat ETM+ data, the panchromatic band is sensitive to EMR in the range of 0.52-0.90 micrometres. This corresponds closely to the green (band 2), red (band 3), and near-infrared (band 4).

Reference:

Jensen, J. R. (2015). Introductory Digital Image Processing: A Remote Sensing Perspective.

See Also: CreateColourComposite

Parameters:

Python function:

wbt.panchromatic_sharpening(
    pan, 
    output, 
    red=None, 
    green=None, 
    blue=None, 
    composite=None, 
    method="brovey", 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=PanchromaticSharpening -v ^
--wd="/path/to/data/" --red=red.tif --green=green.tif ^
--blue=blue.tif --pan=pan.tif --output=pan_sharp.tif ^
--method='brovey'
>>./whitebox_tools -r=PanchromaticSharpening ^
-v --wd="/path/to/data/" --composite=image.tif --pan=pan.tif ^
--output=pan_sharp.tif --method='ihs' 

Source code on GitHub

Author: Dr. John Lindsay

Created: 27/07/2017

Last Modified: 11/02/2019

PercentageContrastStretch

This tool performs a percentage contrast stretch on a raster image. This operation maps each grid cell value in the input raster image (z_in) onto a new scale that ranges from a lower-tail clip value (min_val) to the upper-tail clip value (max_val), with the user-specified number of tonal values (num_tones), such that:

z_out = ((z_in – min_val)/(max_val – min_val)) x num_tones

where z_out is the output value. The values of min_val and max_val are determined from the frequency distribution and the user-specified tail clip value (--clip). For example, if a value of 1% is specified, the tool will determine the values in the input image for which 1% of the grid cells have a lower value min_val and 1% of the grid cells have a higher value max_val. The user must also specify which tails (upper, lower, or both) to clip (--tail).

This is a type of linear contrast stretch with saturation at the tails of the frequency distribution. This is the same kind of stretch that is used to display raster type data on the fly in many GIS software packages, such that the lower and upper tail values are set using the minimum and maximum display values and the number of tonal values is determined by the number of palette entries.

See Also: GaussianContrastStretch, HistogramEqualization, MinMaxContrastStretch, SigmoidalContrastStretch, StandardDeviationContrastStretch

Parameters:

Python function:

wbt.percentage_contrast_stretch(
    i, 
    output, 
    clip=1.0, 
    tail="both", 
    num_tones=256, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=PercentageContrastStretch -v ^
--wd="/path/to/data/" -i=input.tif -o=output.tif --clip=2.0 ^
--tail='both' --num_tones=1024 

Source code on GitHub

Author: Dr. John Lindsay

Created: 13/07/2017

Last Modified: 30/01/2020

SigmoidalContrastStretch

This tool performs a sigmoidal stretch on a raster image. This is a transformation where the input image value for a grid cell (z_in) is transformed to an output value zout such that:

z_out = (1.0 / (1.0 + exp(gain(cutoff - z))) - a ) / b x num_tones

where,

z = (z_in - MIN) / RANGE,

a = 1.0 / (1.0 + exp(gain x cutoff)),

b = 1.0 / (1.0 + exp(gain x (cutoff - 1.0))) - 1.0 / (1.0 + exp(gain x cutoff)),

MIN and RANGE are the minimum value and data range in the input image respectively and gain and cutoff are user specified parameters (--gain, --cutoff).

Like all of WhiteboxTools's contrast enhancement tools, this operation will work on either greyscale or RGB input images.

See Also: GaussianContrastStretch, HistogramEqualization, MinMaxContrastStretch, PercentageContrastStretch, StandardDeviationContrastStretch

Parameters:

Python function:

wbt.sigmoidal_contrast_stretch(
    i, 
    output, 
    cutoff=0.0, 
    gain=1.0, 
    num_tones=256, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=SigmoidalContrastStretch -v ^
--wd="/path/to/data/" -i=input.tif -o=output.tif --cutoff=0.1 ^
--gain=2.0 --num_tones=1024 

Source code on GitHub

Author: Dr. John Lindsay

Created: 13/07/2017

Last Modified: 30/01/2020

StandardDeviationContrastStretch

This tool performs a standard deviation contrast stretch on a raster image. This operation maps each grid cell value in the input raster image (z_in) onto a new scale that ranges from a lower-tail clip value (min_val) to the upper-tail clip value (max_val), with the user-specified number of tonal values (num_tones), such that:

z_out = ((z_in – min_val)/(max_val – min_val)) x num_tones

where z_out is the output value. The values of min_val and max_val are determined based on the image mean and standard deviation. Specifically, the user must specify the number of standard deviations (--clip or --stdev) to be used in determining the min and max clip values. The tool will then calculate the input image mean and standard deviation and estimate the clip values from these statistics.

This is the same kind of stretch that is used to display raster type data on the fly in many GIS software packages.

See Also: GaussianContrastStretch, HistogramEqualization, MinMaxContrastStretch, PercentageContrastStretch, SigmoidalContrastStretch

Parameters:

Python function:

wbt.standard_deviation_contrast_stretch(
    i, 
    output, 
    stdev=2.0, 
    num_tones=256, 
    callback=default_callback
)

Command-line Interface:

>>./whitebox_tools -r=StandardDeviationContrastStretch -v ^
--wd="/path/to/data/" -i=input.tif -o=output.tif --stdev=2.0 ^
--num_tones=1024 

Source code on GitHub

Author: Dr. John Lindsay

Created: 13/07/2017

Last Modified: 30/01/2020