ImFusion::ImageProcessing Namespace Reference

Low-level algorithms for image processing. More...

Detailed Description

Low-level algorithms for image processing.

Depending on the kind of processing algorithm, different overload sets are offered:

• Algorithms that do not change the ImageDescriptor and can be implemented in-place offer only a single computeXYZ() variant, e.g. computeFlipped().
• Algorithms that change the ImageDescriptor will never work in-place and offer two variants:
  • The createXYZ() variant will allocate a new image to store the result of the image processing algorithm.
  • The computeXYZ() variant will accept an additional output image that must already be allocated by the caller. This allows to minimize the number of needed allocations when the output image can be reused such as in streaming use cases for instance.

Algorithms that create output images of the same type than the input images offer two overloads: A strong TypedImage<T> overload as well as an overload taking the MemImage base class.

Enumerations
enum class	Axis { X = 0 , Y = 1 , Z = 2 }
	Enumeration of image axes for processing options. More...
enum class	ColorWeighting { Average = 0 , LumaBT601 = 1 , LumaBT709 = 2 }
	Enumeration for the methods used for converting color (3 channels) to grayscale image. More...
enum class	BayerPattern { BGGR = 0 , GBRG = 1 , GRGB = 2 , RGGB = 3 }
	Enumeration of bayer patterns https://docs.opencv.org/3.4/de/d25/imgproc_color_conversions.html.
enum	ThresholdingOptions {   AcceptAboveOrEqual = 1 << 0 , AcceptBelow = 1 << 1 , AcceptInside = 1 << 2 , AcceptOutside = 1 << 3 ,   UseStorageValues = 1 << 4 , UseOriginalValues = 1 << 5 , MergeChannels = 1 << 6 }
	Bitmask of options that can be passed to createThresholded() and computeThresholded(). More...
enum class	Normalization { None = 0 , TypeRange = 1 , ValueRange = 2 , ApplyShiftAndScale = 3 }
	List of available normlization options for createFloat(). More...
enum class	ValueDomain { StorageValues , OriginalValues }
	Enumeration of available pixel value domains, cf. Pixel Value Domains. More...

Functions
bool	copyPixelData (const MemImage &src, const vec3i &srcOffset, MemImage &dst, const vec3i &dstOffset, const vec3i &size)
	Perform a copy of continuous raw image data between two images using all channels.
bool	copyPixelData (const MemImage &src, const vec3i &srcPixelOffset, int srcChannelOffset, MemImage &dst, const vec3i &dstPixelOffset, int dstChannelOffset, const vec3i &size, int numChannels)
	Perform a copy of continuous raw image data between two images that can have different numbers of channels.
std::unique_ptr< TypedImage< unsigned char > >	createRGBFromBayerPattern (MemImage &image, BayerPattern bayerPattern)
	convert Bayer Pattern Image to a color image
void	computeFlipped (MemImage &image, Axis flipAxis)
	Flip the given image in-place.
void	computeInverted (MemImage &image, bool useImageRange)
	Invert the pixel intensities of the given image in-place.
void	computeClamped (MemImage &image, const vec2 &range, ValueDomain valueDomain)
	Clamp the image intensity of all pixels to the given range (in-place).
void	swapEndianness (MemImage &image)
	Reverses the order of bytes for all pixels in the image pixel buffer.

Rotate images around their main axes
template<typename T>
std::unique_ptr< TypedImage< T > >	createRotated (const TypedImage< T > &input, Axis rotationAxis, int angle, Axis flipAxis=Axis(-1))
	Create a rotated version of the input image.
std::unique_ptr< MemImage >	createRotated (const MemImage &input, Axis rotationAxis, int angle, Axis flipAxis=Axis(-1))
	Create a rotated version of the input image.
bool	computeRotated (const MemImage &input, MemImage &output, Axis rotationAxis, int angle, Axis flipAxis=Axis(-1))
	Create a rotated version of the input image.
bool	isOrthogonalRotationMatrix (const mat3 &rotationMatrix, double tolerance=0.0)
	Test whether the given matrix forms an orthogonal rotation matrix that can be passed to the corresponding overloads of createRotated() and computeRotated().
template<typename T>
std::unique_ptr< TypedImage< T > >	createRotated (const TypedImage< T > &input, const mat3 &rotationMatrix, double tolerance=0.0)
	Create a rotated version of the input image.
std::unique_ptr< MemImage >	createRotated (const MemImage &input, const mat3 &rotationMatrix, double tolerance=0.0)
	Create a rotated version of the input image.
bool	computeRotated (const MemImage &input, MemImage &output, const mat3 &rotationMatrix, double tolerance=0.0)
	Create a rotated version of the input image.

Crop/pad images
template<typename T>
std::unique_ptr< TypedImage< T > >	createCropped (const TypedImage< T > &input, vec3i newSize, vec3i offset=vec3i(-1, -1, -1))
	Create a cropped version of the input image.
std::unique_ptr< MemImage >	createCropped (const MemImage &input, const vec3i &newSize, const vec3i &offset)
	Create a cropped version of the input image.
bool	computeCropped (const MemImage &input, MemImage &output, const vec3i &offset=vec3i(-1, -1, -1))
	Create a cropped version of the input image.
template<typename T>
std::unique_ptr< TypedImage< T > >	createPadded (const TypedImage< T > &input, PaddingMode paddingMode, const vec3i &paddingLLF, const vec3i &paddingURB, bool legacyMirrorPadding=true)
	Create a padded version of the input image where the additional border pixels have a specified value.
std::unique_ptr< MemImage >	createPadded (const MemImage &input, PaddingMode paddingMode, const vec3i &paddingLLF, const vec3i &paddingURB, bool legacyMirrorPadding=true)
	Create a padded version of the input image where the additional border pixels have a specified value.
bool	computePadded (const MemImage &input, MemImage &output, PaddingMode paddingMode, const vec3i &paddingLLF, const vec3i &paddingURB, bool legacyMirrorPadding=true)
	Create a padded version of the input image where the additional border pixels have a specified value.

Resample images
template<typename T>
std::unique_ptr< TypedImage< T > >	createDownsampled (const TypedImage< T > &input, int dx, int dy, int dz=1, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average)
	Create a downsampled version of the input image, where the new image size is specified by an integer factor.
std::unique_ptr< MemImage >	createDownsampled (const MemImage &input, int dx, int dy, int dz=1, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average)
	Create a downsampled version of the input image, where the new image size is specified by an integer factor.
bool	computeDownsampled (const MemImage &input, MemImage &output, int dx, int dy, int dz=1, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average)
	Create a downsampled version of the input image, where the new image size is specified by an integer factor.
template<typename T>
std::unique_ptr< TypedImage< T > >	createResampledToSpacing (const TypedImage< T > &input, SpacingMode spacingAdjustment, const vec3 &spacing, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average, InterpolationMode interpolationMode=InterpolationMode::Linear, bool allowDimensionChange=false)
	Create a resampled version of the input image with a dedicated target spacing.
std::unique_ptr< MemImage >	createResampledToSpacing (const MemImage &input, SpacingMode spacingAdjustment, const vec3 &spacing, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average, InterpolationMode interpolationMode=InterpolationMode::Linear, bool allowDimensionChange=false)
	Create a resampled version of the input image with a dedicated target spacing.
bool	computeResampledToSpacing (const MemImage &input, MemImage &output, SpacingMode spacingAdjustment, const vec3 &spacing, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average, InterpolationMode interpolationMode=InterpolationMode::Linear, bool allowDimensionChange=false)
	Create a resampled version of the input image with a dedicated target spacing.
template<typename T>
std::unique_ptr< TypedImage< T > >	createResampledToDimensions (const TypedImage< T > &input, const vec3i &dimensions, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average, InterpolationMode interpolationMode=InterpolationMode::Linear)
	Create a resampled version of the input image with a dedicated target resolution.
std::unique_ptr< MemImage >	createResampledToDimensions (const MemImage &input, const vec3i &dimensions, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average, InterpolationMode interpolationMode=InterpolationMode::Linear)
	Create a resampled version of the input image with a dedicated target resolution.
bool	computeResampledToDimensions (const MemImage &input, MemImage &output, const vec3i &dimensions, bool zeroMask=false, ReductionMode reductionMode=ReductionMode::Average, InterpolationMode interpolationMode=InterpolationMode::Linear)
	Create a resampled version of the input image with a dedicated target resolution.

Binary thresholding based on pixel values
std::unique_ptr< TypedImage< unsigned char > >	createThresholded (const MemImage &input, double value, Flags< ThresholdingOptions > options)
	Create a binary mask image based on the input image's pixel values.
bool	computeThresholded (const MemImage &input, TypedImage< unsigned char > &output, double value, Flags< ThresholdingOptions > options)
	Create a binary mask image based on the input image's pixel values.
std::unique_ptr< TypedImage< unsigned char > >	createThresholded (const MemImage &input, vec2 interval, Flags< ThresholdingOptions > options)
	Create a binary mask image based on the input image's pixel values.
bool	computeThresholded (const MemImage &input, TypedImage< unsigned char > &output, vec2 interval, Flags< ThresholdingOptions > options)
	Create a binary mask image based on the input image's pixel values.

Change the pixel representation/format
template<typename InputT, typename OutputT>
std::unique_ptr< TypedImage< OutputT > >	createConverted (const TypedImage< InputT > &input, std::optional< std::pair< InputT, InputT > > inputRange=std::nullopt, std::optional< std::pair< OutputT, OutputT > > outputRange=std::nullopt)
	Performs a type conversion from a TypedImage<InputT> to a newly created TypedImage<OutputT>.
template<typename OutputT>
std::unique_ptr< TypedImage< OutputT > >	createConverted (const MemImage &input, std::optional< std::pair< double, double > > inputRange=std::nullopt, std::optional< std::pair< OutputT, OutputT > > outputRange=std::nullopt)
	Performs a type conversion from any MemImage to a newly created TypedImage<OutputT>.
std::unique_ptr< TypedImage< float > >	createFloat (const MemImage &input, Normalization valueMapping)
	Create a floating point version of the input image.
bool	createFloat (const MemImage &input, TypedImage< float > &output, Normalization valueMapping)
	Create a floating point version of the input image.
template<typename T>
std::unique_ptr< TypedImage< T > >	createGrayscale (const TypedImage< T > &input, int numChannels=-1, ColorWeighting method=ColorWeighting::Average)
	Create a single-channel version of the input image by averaging pixel values over the first N (or all if -1) channels.
bool	computeGrayscale (const MemImage &input, MemImage &output, int numChannels=-1, ColorWeighting method=ColorWeighting::Average)
	Create a single-channel version of the input image by averaging pixel values over first N (or all if -1) channels.
std::unique_ptr< MemImage >	createGrayscale (const MemImage &input, int numChannels=-1, ColorWeighting method=ColorWeighting::Average)
	Create a single-channel version of the input image by averaging pixel values over first N (or all if -1) channels.

Compute gradients
template<typename T>
void	computeGradientMagnitude (const TypedImage< T > *img, TypedImage< float > *gradient, bool normalizeResult=true, TypedImage< float > *orientation=0)
	Compute normalized gradient magnitude using 3x3 Sobel filter.
template<typename T>
void	computeGradient (const TypedImage< T > *img, TypedImage< float > *gx, TypedImage< float > *gy, TypedImage< float > *gNorm=0)
template<typename T>
void	computeGradient2D (const TypedImage< T > *img, TypedImage< float > *gx, TypedImage< float > *gy)
template<typename T>
void	computeGradient3D (const TypedImage< T > *img, TypedImage< float > *gx, TypedImage< float > *gy, TypedImage< float > *gz)
template<typename T>
std::unique_ptr< TypedImage< T > >	addOrRemoveAlphaChannel (const TypedImage< T > &img)
std::unique_ptr< MemImage >	addOrRemoveAlphaChannel (const MemImage &input)
void	computeSubpixelEdgeLocationAndOrientation2D (const TypedImage< float > *g, const TypedImage< float > *gx, const TypedImage< float > *gy, float threshold, std::vector< std::pair< vec2, vec2 > > &location)
void	computeSubpixelEdgeLocationAndOrientation3D (const TypedImage< float > *g, const TypedImage< float > *gx, const TypedImage< float > *gy, const TypedImage< float > *gz, float threshold, std::vector< std::pair< vec3, vec3 > > &location)

Enumeration Type Documentation

Axis

enum class Axis

strong

Enumeration of image axes for processing options.

Enumerator
X	X axis.
Y	Y axis.
Z	Z axis.

ColorWeighting

enum class ColorWeighting

strong

Enumeration for the methods used for converting color (3 channels) to grayscale image.

if the number of channels considered during the conversion is different to 3, then the Average is used independent of the user choice

Note

the problems of color averaging method can be found in: https://www.tutorialspoint.com/dip/grayscale_to_rgb_conversion.htm

opencv uses LumaBT601: https://docs.opencv.org/3.4/de/d25/imgproc_color_conversions.html

for more info about the difference between BT601 and BT709: https://en.wikipedia.org/wiki/Luma_(video)

Enumerator
Average	Grayscale = R / 3 + G / 3 + B / 3.
LumaBT601	Grayscale = 0.299R + 0.587G + 0.114B (used in opencv and most digital standard definition formats)
LumaBT709	Grayscale = 0.2126R + 0.7152G + 0.0722B.

ThresholdingOptions

enum ThresholdingOptions

Bitmask of options that can be passed to createThresholded() and computeThresholded().

Note

Not all combinations of options are valid.

Enumerator
AcceptAboveOrEqual	All pixels above or equal to the threshold will be marked 1; only valid for the float overload.
AcceptBelow	All pixels below the threshold will be marked 1; only valid for the float overload.
AcceptInside	All pixels inside the closed interval will be marked 1; only valid for the vec2 overload.
AcceptOutside	All pixels outside the closed interval will be marked 1; only valid for the vec2 overload.
UseStorageValues	The value/interval parameter describes storage values.
UseOriginalValues	The value/interval parameter describes original values.
MergeChannels	Compute threshold value based on average intensity of all channels and return a single channel threshold image.

Normalization

enum class Normalization

strong

List of available normlization options for createFloat().

Enumerator
None	Do not perform any normalization, simply cast the pixel value to float.
TypeRange	Use entire type range to normalize values to [0..1] range.
ValueRange	Use minimum/maximum pixel intensity as normalization range;.
ApplyShiftAndScale	Do not perform normalization but apply shift and scale to convert to original values.

ValueDomain

enum class ValueDomain

strong

Enumeration of available pixel value domains, cf. Pixel Value Domains.

Enumerator
StorageValues	Values as stored in memory.
OriginalValues	Values as recorded in the original source if available (e.g. Hounsfield Units for CT images)

Function Documentation

copyPixelData() [1/2]

bool copyPixelData	(	const MemImage &	src,
		const vec3i &	srcOffset,
		MemImage &	dst,
		const vec3i &	dstOffset,
		const vec3i &	size )

Perform a copy of continuous raw image data between two images using all channels.

The image sizes do not need to match, you can specify any subregion in both the source and destination image. If the specified region exceeds any of the input/output images' dimensions it will be clamped to fit. Both the image type and the number of channels of src and dst must match.

Parameters

src	Source image to copy data from
srcOffset	Pixel offset in source subvolume to copy pixel data from
dst	Destination image to copy data to
dstOffset	Pixel offset in destination subvolume to copy pixel data to
size	Size of subvolume to copy

copyPixelData() [2/2]

bool copyPixelData	(	const MemImage &	src,
		const vec3i &	srcPixelOffset,
		int	srcChannelOffset,
		MemImage &	dst,
		const vec3i &	dstPixelOffset,
		int	dstChannelOffset,
		const vec3i &	size,
		int	numChannels )

Perform a copy of continuous raw image data between two images that can have different numbers of channels.

Neither the image sizes nor the number of channels do need to match, you can specify any subregion in both the source and destination image. If the specified regions exceed any of the input/output images' dimensions/number of channels they will be clamped to fit. The image type of src and dst must match.

Parameters

src	Source image to copy data from
srcPixelOffset	Pixel offset in source subvolume to copy pixel data from
srcChannelOffset	First channel in source subvolume to copy channel data from for each pixel, must be >= 0
dst	Destination image to copy data to
dstPixelOffset	Pixel offset in destination subvolume to copy pixel data to
dstChannelOffset	First channel in destination subvolume to copy channel data to for each pixel, must be >= 0
size	Size of subvolume to copy
numChannels	Number of channels to copy for each pixel

createRotated() [1/4]

template<typename T>

std::unique_ptr< TypedImage< T > > createRotated	(	const TypedImage< T > &	input,
		Axis	rotationAxis,
		int	angle,
		Axis	flipAxis = Axis(-1) )

Create a rotated version of the input image.

This function does only support orthogonal rotations.

Parameters

input	Input image, will not be modified
rotationAxis	Rotation axis to rotate around
angle	Rotation angle in degrees, must be a multiple of 90. A positive angle refers to a counter-clockwise rotation around the rotation axis. For more information about the coordinate system conventions see Coordinate Systems.
flipAxis	Optional axis to perform an additional flip of the pixel data in.

createRotated() [2/4]

std::unique_ptr< MemImage > createRotated	(	const MemImage &	input,
		Axis	rotationAxis,
		int	angle,
		Axis	flipAxis = Axis(-1) )

Create a rotated version of the input image.

This function only supports orthogonal rotations.

Parameters

input	Input image, will not be modified
rotationAxis	Rotation axis to rotate around
angle	Rotation angle in degrees, must be a multiple of 90 within the interval [-270, 270]. A positive angle refers to a counter-clockwise rotation around the rotation axis. For more information about the coordinate system conventions see Coordinate Systems.
flipAxis	Optional axis to perform an additional flip of the pixel data.

computeRotated() [1/2]

bool computeRotated	(	const MemImage &	input,
		MemImage &	output,
		Axis	rotationAxis,
		int	angle,
		Axis	flipAxis = Axis(-1) )

Create a rotated version of the input image.

This function only supports orthogonal rotations.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with the processing applied to the input image. In case it's not the function returns false.
rotationAxis	Rotation axis to rotate around
angle	Rotation angle in degrees, must be a multiple of 90 within the interval [-270, 270]. A positive angle refers to a counter-clockwise rotation around the rotation axis. For more information about the coordinate system conventions see Coordinate Systems.
flipAxis	Optional axis to perform an additional flip of the pixel data.

isOrthogonalRotationMatrix()

bool isOrthogonalRotationMatrix	(	const mat3 &	rotationMatrix,
		double	tolerance = 0.0 )

Test whether the given matrix forms an orthogonal rotation matrix that can be passed to the corresponding overloads of createRotated() and computeRotated().

Parameters

rotationMatrix	Rotation matrix to decompose, will fail if not orthogonal.
tolerance	Maximum allowable deviation of the euler angles from an orthogonal configuration in degrees.

createRotated() [3/4]

template<typename T>

std::unique_ptr< TypedImage< T > > createRotated	(	const TypedImage< T > &	input,
		const mat3 &	rotationMatrix,
		double	tolerance = 0.0 )

Create a rotated version of the input image.

This function does only support 3D volumes and orthogonal rotations.

Parameters

input	Input image, will not be modified
rotationMatrix	Rotation matrix to apply, must be orthogonal.
tolerance	Maximum allowable deviation of the euler angles from an orthogonal configuration in degrees.

createRotated() [4/4]

std::unique_ptr< MemImage > createRotated	(	const MemImage &	input,
		const mat3 &	rotationMatrix,
		double	tolerance = 0.0 )

Create a rotated version of the input image.

This function does only support 3D volumes and orthogonal rotations.

Parameters

input	Input image, will not be modified
rotationMatrix	Rotation matrix to apply, must be orthogonal.
tolerance	Maximum allowable deviation of the euler angles from an orthogonal configuration in degrees.

computeRotated() [2/2]

bool computeRotated	(	const MemImage &	input,
		MemImage &	output,
		const mat3 &	rotationMatrix,
		double	tolerance = 0.0 )

Create a rotated version of the input image.

This function does only support 3D volumes and orthogonal rotations.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with the processing applied to the input image. In case it's not the function returns false.
rotationMatrix	Rotation matrix to apply, must be orthogonal.
tolerance	Maximum allowable deviation of the euler angles from an orthogonal configuration in degrees.

createCropped() [1/2]

template<typename T>

std::unique_ptr< TypedImage< T > > createCropped	(	const TypedImage< T > &	input,
		vec3i	newSize,
		vec3i	offset = vec3i(-1, -1, -1) )

Create a cropped version of the input image.

Parameters

input	Input image, will not be modified
newSize	Size of the cropped image
offset	Pixel offset of cropping region in input image. An offset of -1 will keep put the cropping region in the center of the corresponding axis.

createCropped() [2/2]

std::unique_ptr< MemImage > createCropped	(	const MemImage &	input,
		const vec3i &	newSize,
		const vec3i &	offset )

Create a cropped version of the input image.

Parameters

input	Input image, will not be modified
newSize	Size of the cropped image
offset	Pixel offset of cropping region in input image. An offset of -1 will keep put the cropping region in the center of the corresponding axis.

computeCropped()

bool computeCropped	(	const MemImage &	input,
		MemImage &	output,
		const vec3i &	offset = vec3i(-1, -1, -1) )

Create a cropped version of the input image.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor is used for the cropping dimensions.
offset	Pixel offset of cropping region in input image. An offset of -1 will keep put the cropping region in the center of the corresponding axis.

createPadded() [1/2]

template<typename T>

std::unique_ptr< TypedImage< T > > createPadded	(	const TypedImage< T > &	input,
		PaddingMode	paddingMode,
		const vec3i &	paddingLLF,
		const vec3i &	paddingURB,
		bool	legacyMirrorPadding = true )

Create a padded version of the input image where the additional border pixels have a specified value.

Parameters

input	Input image, will not be modified
paddingMode	Padding mode specifying what pixel values to assign to the additional border pixels
paddingLLF	Number of padding pixels in X, Y, and Z direction at the lower left front corner.
paddingURB	Number of padding pixels in X, Y, and Z direction at the upper right back corner.
legacyMirrorPadding	Before SUITE-3748, mirror padding was not computed correctly on URB side. This flag allows to continue use the wrong definition of mirror padding to avoid regressions. Note: this is a temporary parameter and will be deprecated/removed soon.

createPadded() [2/2]

std::unique_ptr< MemImage > createPadded	(	const MemImage &	input,
		PaddingMode	paddingMode,
		const vec3i &	paddingLLF,
		const vec3i &	paddingURB,
		bool	legacyMirrorPadding = true )

Create a padded version of the input image where the additional border pixels have a specified value.

Parameters

input	Input image, will not be modified
paddingMode	Padding mode specifying what pixel values to assign to the additional border pixels
paddingLLF	Number of padding pixels in X, Y, and Z direction at the lower left front corner.
paddingURB	Number of padding pixels in X, Y, and Z direction at the upper right back corner.
legacyMirrorPadding	Before SUITE-3748, mirror padding was not computed correctly on URB side. This flag allows to continue use the wrong definition of mirror padding to avoid regressions. Note: this is a temporary parameter and will be deprecated/removed soon.

computePadded()

bool computePadded	(	const MemImage &	input,
		MemImage &	output,
		PaddingMode	paddingMode,
		const vec3i &	paddingLLF,
		const vec3i &	paddingURB,
		bool	legacyMirrorPadding = true )

Create a padded version of the input image where the additional border pixels have a specified value.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with the processing applied to the input image. In case it's not the function returns false.
paddingMode	Padding mode specifying what pixel values to assign to the additional border pixels
paddingLLF	Number of padding pixels in X, Y, and Z direction at the lower left front corner.
paddingURB	Number of padding pixels in X, Y, and Z direction at the upper right back corner.
legacyMirrorPadding	Before SUITE-3748, mirror padding was not computed correctly on URB side. This flag allows to continue use the wrong definition of mirror padding to avoid regressions. Note: this is a temporary parameter and will be deprecated/removed soon.

createDownsampled() [1/2]

template<typename T>

std::unique_ptr< TypedImage< T > > createDownsampled	(	const TypedImage< T > &	input,
		int	dx,
		int	dy,
		int	dz = 1,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average )

Create a downsampled version of the input image, where the new image size is specified by an integer factor.

Parameters

input	Input image, will not be modified
dx	Downsampling factor in X direction
dy	Downsampling factor in Y direction
dz	Downsampling factor in Z direction
zeroMask	If enabled zeros are propagated to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value

createDownsampled() [2/2]

std::unique_ptr< MemImage > createDownsampled	(	const MemImage &	input,
		int	dx,
		int	dy,
		int	dz = 1,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average )

Create a downsampled version of the input image, where the new image size is specified by an integer factor.

Parameters

input	Input image, will not be modified
dx	Downsampling factor in X direction
dy	Downsampling factor in Y direction
dz	Downsampling factor in Z direction
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value

computeDownsampled()

bool computeDownsampled	(	const MemImage &	input,
		MemImage &	output,
		int	dx,
		int	dy,
		int	dz = 1,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average )

Create a downsampled version of the input image, where the new image size is specified by an integer factor.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with the processing applied to the input image. In case it's not the function returns false.
dx	Downsampling factor in X direction
dy	Downsampling factor in Y direction
dz	Downsampling factor in Z direction
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value

createResampledToSpacing() [1/2]

template<typename T>

std::unique_ptr< TypedImage< T > > createResampledToSpacing	(	const TypedImage< T > &	input,
		SpacingMode	spacingAdjustment,
		const vec3 &	spacing,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average,
		InterpolationMode	interpolationMode = InterpolationMode::Linear,
		bool	allowDimensionChange = false )

Create a resampled version of the input image with a dedicated target spacing.

Parameters

input	Input image, will not be modified
spacingAdjustment	If set to ADJUST, resampling is performed to roughly the desired spacing so that the image extent is kept constant. Otherwise, the exact spacing defined will be used, leading to a potentially different extent.
spacing	Target spacing of the resampled image
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value
interpolationMode	Interpolation mode for pixel lookup in input image
allowDimensionChange	If true, 3D images may become 2D images, etc.

createResampledToSpacing() [2/2]

std::unique_ptr< MemImage > createResampledToSpacing	(	const MemImage &	input,
		SpacingMode	spacingAdjustment,
		const vec3 &	spacing,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average,
		InterpolationMode	interpolationMode = InterpolationMode::Linear,
		bool	allowDimensionChange = false )

Create a resampled version of the input image with a dedicated target spacing.

Parameters

input	Input image, will not be modified
spacingAdjustment	If set to ADJUST, resampling is performed to roughly the desired spacing so that the image extent is kept constant. Otherwise, the exact spacing defined will be used, leading to a potentially different extent.
spacing	Target spacing of the resampled image
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value
interpolationMode	Interpolation mode for pixel lookup in input image
allowDimensionChange	If true, 3D images may become 2D images, etc.

computeResampledToSpacing()

bool computeResampledToSpacing	(	const MemImage &	input,
		MemImage &	output,
		SpacingMode	spacingAdjustment,
		const vec3 &	spacing,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average,
		InterpolationMode	interpolationMode = InterpolationMode::Linear,
		bool	allowDimensionChange = false )

Create a resampled version of the input image with a dedicated target spacing.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with the processing applied to the input image. In case it's not the function returns false.
spacingAdjustment	If set to ADJUST, resampling is performed to roughly the desired spacing so that the image extent is kept constant. Otherwise, the exact spacing defined will be used, leading to a potentially different extent.
spacing	Target spacing of the resampled image
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value
interpolationMode	Interpolation mode for pixel lookup in input image
allowDimensionChange	If true, 3D images may become 2D images, etc.

createResampledToDimensions() [1/2]

template<typename T>

std::unique_ptr< TypedImage< T > > createResampledToDimensions	(	const TypedImage< T > &	input,
		const vec3i &	dimensions,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average,
		InterpolationMode	interpolationMode = InterpolationMode::Linear )

Create a resampled version of the input image with a dedicated target resolution.

Parameters

input	Input image, will not be modified
dimensions	Target pixel dimensions of resampled image
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value
interpolationMode	Interpolation mode for pixel lookup in input image

createResampledToDimensions() [2/2]

std::unique_ptr< MemImage > createResampledToDimensions	(	const MemImage &	input,
		const vec3i &	dimensions,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average,
		InterpolationMode	interpolationMode = InterpolationMode::Linear )

Create a resampled version of the input image with a dedicated target resolution.

Parameters

input	Input image, will not be modified
dimensions	Target pixel dimensions of resampled image
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value
interpolationMode	Interpolation mode for pixel lookup in input image

computeResampledToDimensions()

bool computeResampledToDimensions	(	const MemImage &	input,
		MemImage &	output,
		const vec3i &	dimensions,
		bool	zeroMask = false,
		ReductionMode	reductionMode = ReductionMode::Average,
		InterpolationMode	interpolationMode = InterpolationMode::Linear )

Create a resampled version of the input image with a dedicated target resolution.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with the processing applied to the input image. In case it's not the function returns false.
dimensions	Target pixel dimensions of resampled image
zeroMask	If enabled, zeros are always propagated (independent of reduction mode) to avoid masking artifacts
reductionMode	Filter mode defining how contributing input pixels are combined into the downsampled pixel value
interpolationMode	Interpolation mode for pixel lookup in input image

createThresholded() [1/2]

std::unique_ptr< TypedImage< unsigned char > > createThresholded	(	const MemImage &	input,
		double	value,
		Flags< ThresholdingOptions >	options )

Create a binary mask image based on the input image's pixel values.

Parameters

input	Input image, will not be modified
value	Cutoff value for thresholding, either in storage or in original value domain depending on options.
options	Set of thresholding options, must meet the following conditions Contains either AcceptAboveOrEqual or AcceptBelow Contains either UseStorageValues or UseOriginalValues Does not contain AcceptInside or AcceptOutside Can contain any combination of the other options

computeThresholded() [1/2]

bool computeThresholded	(	const MemImage &	input,
		TypedImage< unsigned char > &	output,
		double	value,
		Flags< ThresholdingOptions >	options )

Create a binary mask image based on the input image's pixel values.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with that of the input image. In case it's not the function returns false.
value	Cutoff value for thresholding, either in storage or in original value domain depending on options.
options	Set of thresholding options, must meet the following conditions Contains either AcceptAboveOrEqual or AcceptBelow Contains either UseStorageValues or UseOriginalValues Does not contain AcceptInside or AcceptOutside Can contain any combination of the other options

createThresholded() [2/2]

std::unique_ptr< TypedImage< unsigned char > > createThresholded	(	const MemImage &	input,
		vec2	interval,
		Flags< ThresholdingOptions >	options )

Create a binary mask image based on the input image's pixel values.

Parameters

input	Input image, will not be modified
interval	Closed interval describing the value range for thresholding, either in storage or in original value domain depending on options.
options	Set of thresholding options, must meet the following conditions Contains either AcceptInside or AcceptOutside Contains either UseStorageValues or UseOriginalValues Does not contain AcceptAboveOrEqual or AcceptBelow Can contain any combination of the other options

computeThresholded() [2/2]

bool computeThresholded	(	const MemImage &	input,
		TypedImage< unsigned char > &	output,
		vec2	interval,
		Flags< ThresholdingOptions >	options )

Create a binary mask image based on the input image's pixel values.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with that of the input image. In case it's not the function returns false.
interval	Closed interval describing the value range for thresholding, either in storage or in original value domain depending on options.
options	Set of thresholding options, must meet the following conditions Contains either AcceptInside or AcceptOutside Contains either UseStorageValues or UseOriginalValues Does not containAcceptAboveOrEqual or AcceptBelow Can contain any combination of the other options

createRGBFromBayerPattern()

std::unique_ptr< TypedImage< unsigned char > > createRGBFromBayerPattern	(	MemImage &	image,
		BayerPattern	bayerPattern )

convert Bayer Pattern Image to a color image

Parameters

image	Bayer pattern Image to be converted to RGB, should be a grayscale unsigned 8 bit image
BayerPattern	the Bayer pattern of the input image

Returns

return a pointer to the result image (8 bit RGB image), nullptr in case of failure

computeFlipped()

void computeFlipped	(	MemImage &	image,
		Axis	flipAxis )

Flip the given image in-place.

Parameters

image	Image to flip
flipAxis	Flip axis

computeInverted()

void computeInverted	(	MemImage &	image,
		bool	useImageRange )

Invert the pixel intensities of the given image in-place.

Parameters

image	Image to invert
useImageRange	If true, the output values will be such that the new image minimum will be equal to the old image maximum (and vice versa). Otherwise the full range of the image type will be used to invert the values.

computeClamped()

void computeClamped	(	MemImage &	image,
		const vec2 &	range,
		ValueDomain	valueDomain )

Clamp the image intensity of all pixels to the given range (in-place).

Parameters

image	Input image, will be changed
range	Value range to which to clamp the image pixels.
valueDomain	Pixel value domain in which range is specified, cf. Pixel Value Domains.

createConverted() [1/2]

template<typename InputT, typename OutputT>

std::unique_ptr< TypedImage< OutputT > > createConverted	(	const TypedImage< InputT > &	input,
		std::optional< std::pair< InputT, InputT > >	inputRange = std::nullopt,
		std::optional< std::pair< OutputT, OutputT > >	outputRange = std::nullopt )

Performs a type conversion from a TypedImage<InputT> to a newly created TypedImage<OutputT>.

The pixel values are adjusted linearly so that inputRange.first is mapped to outputRange.first and inputRange.second is mapped to outputRange.second. Shift and scale of the resulting ImageDescriptor are adjusted accordingly so that the original/real-world values remain the same.

Note

If both InputT and OutputT are integral types and the outputRange is larger than the inputRange this function will truncate any floating-point scaling factor between the two to an integer one in order to avoid a re-quantization of the input values. In this case inputRange.second is mapped to the largest value less or equal than outputRange.second so that outputRange/inputRange is an integer division.

Parameters

input	Input image to convert to a different type.
inputRange	Storage pixel value range of input that should be mapped to outputRange If unset, will use {MIN_T, MAX_T} for integral types and {0.0, 1.0} for floating point types.
outputRange	Storage pixel value range of the return value to which pixels from inputRange are mapped to. If unset, will use {MIN_T, MAX_T} for integral types and {0.0, 1.0} for floating point types.

createConverted() [2/2]

template<typename OutputT>

std::unique_ptr< TypedImage< OutputT > > createConverted	(	const MemImage &	input,
		std::optional< std::pair< double, double > >	inputRange = std::nullopt,
		std::optional< std::pair< OutputT, OutputT > >	outputRange = std::nullopt )

Performs a type conversion from any MemImage to a newly created TypedImage<OutputT>.

The pixel values are adjusted linearly so that inputRange.first is mapped to outputRange.first and inputRange.second is mapped to outputRange.second. Shift and scale of the resulting ImageDescriptor are adjusted accordingly so that the original/real-world values remain the same.

Note

If both InputT and OutputT are integral types and the outputRange is larger than the inputRange this function will truncate any floating-point scaling factor between the two to an integer one in order to avoid a re-quantization of the input values. In this case inputRange.second is mapped to the largest value less or equal than outputRange.second so that outputRange/inputRange it is an integer division.

Parameters

input	Input image to convert to a different type.
inputRange	Storage pixel value range of input that should be mapped to outputRange If unset, will use {MIN_T, MAX_T} for integral types and {0.0, 1.0} for floating point types.
outputRange	Storage pixel value range of the return value to which pixels from inputRange are mapped to. If unset, will use {MIN_T, MAX_T} for integral types and {0.0, 1.0} for floating point types.

createFloat() [1/2]

std::unique_ptr< TypedImage< float > > createFloat	(	const MemImage &	input,
		Normalization	valueMapping )

Create a floating point version of the input image.

Parameters

input	Input image, will not be modified
valueMapping	Defines how input pixel intensities are mapped to float

createFloat() [2/2]

bool createFloat	(	const MemImage &	input,
		TypedImage< float > &	output,
		Normalization	valueMapping )

Create a floating point version of the input image.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with that of the input image. In case it's not the function returns false.
valueMapping	Defines how input pixel intensities are mapped to float

createGrayscale() [1/2]

template<typename T>

std::unique_ptr< TypedImage< T > > createGrayscale	(	const TypedImage< T > &	input,
		int	numChannels = -1,
		ColorWeighting	method = ColorWeighting::Average )

Create a single-channel version of the input image by averaging pixel values over the first N (or all if -1) channels.

Parameters

input	Input image, will not be modified
numChannels	Number of channels to be considered
method	The method used for computing the grayscale values, fall back to Averaging if the number of channels is != 3

computeGrayscale()

bool computeGrayscale	(	const MemImage &	input,
		MemImage &	output,
		int	numChannels = -1,
		ColorWeighting	method = ColorWeighting::Average )

Create a single-channel version of the input image by averaging pixel values over first N (or all if -1) channels.

Parameters

input	Input image, will not be modified
output	User provided output image, its descriptor must be compatible with the processing applied to the input image. In case it's not the function returns false.
numChannels	Number of channels to be considered
method	The method used for computing the grayscale values, fall back to Averaging if the number of channels is != 3

createGrayscale() [2/2]

std::unique_ptr< MemImage > createGrayscale	(	const MemImage &	input,
		int	numChannels = -1,
		ColorWeighting	method = ColorWeighting::Average )

Create a single-channel version of the input image by averaging pixel values over first N (or all if -1) channels.

Parameters

input	Input image, will not be modified
numChannels	Number of channels to be considered
method	The method used for computing the grayscale values, fall back to Averaging if the number of channels is != 3

computeGradientMagnitude()

template<typename T>

void computeGradientMagnitude	(	const TypedImage< T > *	img,
		TypedImage< float > *	gradient,
		bool	normalizeResult = true,
		TypedImage< float > *	orientation = 0 )

Compute normalized gradient magnitude using 3x3 Sobel filter.

The border of the image is set to zero.

Parameters

	img	Input image
[out]	gradient	Gradient image. Has to be float type.
	normalizeResult	Normalize result to range [0;1]
[out]	orientation	Optional pointer to orientation image receiving gradient orientation