Classes
struct	stbir_kernel_and_support

Enumerations
enum class	EResizeImageFilter : uint32 { Default = 0 , PointSample , Box , Triangle , Bilinear = Triangle , CubicGaussian , CubicSharp , CubicMitchell , AdaptiveSharp , AdaptiveSmooth , MitchellOneQuarter , MitchellOneSixth , MitchellNegOneSixth , MitchellNegOneThird , Lanczos4 , Lanczos5 , WithoutFlagsMask = 63 , Flag_WrapX = 64 , Flag_WrapY = 128 , Flag_AlphaWeighted = 256 }

enum class	ProcessLinearPixelsAction { ReadOnly , Modified }

Functions
IMAGECORE_API void	CopyImage (const FImageView &SrcImage, const FImageView &DestImage)

IMAGECORE_API void	CopyImageRGBABGRA (const FImageView &SrcImage, const FImageView &DestImage)

IMAGECORE_API void	CopyImageTo2U16 (const FImageView &SrcImage, const FImageView &DestImage)

IMAGECORE_API void	TransposeImageRGBABGRA (const FImageView &Image)

IMAGECORE_API void	SanitizeFloat16AndSetAlphaOpaqueForBC6H (const FImageView &InOutImage)

IMAGECORE_API bool	DetectAlphaChannel (const FImageView &InImage)

IMAGECORE_API void	SetAlphaOpaque (const FImageView &InImage)

IMAGECORE_API void	ResizeTo (const FImageView &SourceImage, FImage &DestImage, int32 DestSizeX, int32 DestSizeY, ERawImageFormat::Type DestFormat, EGammaSpace DestGammaSpace)

IMAGECORE_API void	ComputeChannelLinearMinMax (const FImageView &InImage, FLinearColor &OutMin, FLinearColor &OutMax)

IMAGECORE_API bool	ScaleChannelsSoMinMaxIsInZeroToOne (const FImageView &ImageToModify)

IMAGECORE_API FLinearColor	ComputeImageLinearAverage (const FImageView &Image)

IMAGECORE_API void	TransformToWorkingColorSpace (const FImageView &InLinearImage, const FVector2d &SourceRedChromaticity, const FVector2d &SourceGreenChromaticity, const FVector2d &SourceBlueChromaticity, const FVector2d &SourceWhiteChromaticity, UE::Color::EChromaticAdaptationMethod Method, double EqualityTolerance=1.e-7)

	ENUM_CLASS_FLAGS (EResizeImageFilter)

IMAGECORE_API void	ResizeImage (const FImageView &SourceImage, const FImageView &DestImage, EResizeImageFilter Filter=EResizeImageFilter::Default)

IMAGECORE_API void	ResizeImageAllocDest (const FImageView &SourceImage, FImage &DestImage, int32 DestSizeX, int32 DestSizeY, ERawImageFormat::Type DestFormat, EGammaSpace DestGammaSpace, EResizeImageFilter Filter=EResizeImageFilter::Default)

IMAGECORE_API void	ResizeImageAllocDest (const FImageView &SourceImage, FImage &DestImage, int32 DestSizeX, int32 DestSizeY, EResizeImageFilter Filter=EResizeImageFilter::Default)

IMAGECORE_API void	ResizeImageInPlace (FImage &Image, int32 DestSizeX, int32 DestSizeY, ERawImageFormat::Type DestFormat, EGammaSpace DestGammaSpace, EResizeImageFilter Filter=EResizeImageFilter::Default)

IMAGECORE_API void	ResizeImageInPlace (FImage &Image, int32 DestSizeX, int32 DestSizeY, EResizeImageFilter Filter=EResizeImageFilter::Default)

int64	ImageParallelForComputeNumRows (const FImageView &Image)

FImageView	ImageParallelForGetOneRowView (const FImageView &Image, int64 Y)

IMAGECORE_API int32	ImageParallelForComputeNumJobs (const FImageView &Image, int64 *pRowsPerJob)

IMAGECORE_API int64	ImageParallelForMakePart (FImageView *Part, const FImageView &Whole, int64 JobIndex, int64 RowsPerJob)

template<typename Lambda >
void	ImageParallelFor (const TCHAR *DebugName, const FImageView &Image, const Lambda &Func)

template<typename Lambda >
void	ProcessLinearPixels (const FImageView &Image, const Lambda &Func, int64 StartY)

template<typename Lambda >
void	ImageParallelProcessLinearPixels (const TCHAR *DebugName, const FImageView &Image, const Lambda &Func)

Detailed Description

The easy and efficient way to write a pixel processing loop that supports all pixel formats is to use

ImageParallelProcessLinearPixels

passing in a lambda like : ProcessLinearPixelsAction ProcessPixels(TArrayView64<FLinearColor> Colors)

this will let you visit all the pixels as FLinearColor.

You can use this for read (return ProcessLinearPixelsAction::ReadOnly) or read-write (return ProcessLinearPixelsAction::Modified).

Enumeration Type Documentation

◆ EResizeImageFilter

enum class FImageCore::EResizeImageFilter : uint32

strong

Enumerator
Default
PointSample
Box
Triangle
Bilinear
CubicGaussian
CubicSharp
CubicMitchell
AdaptiveSharp
AdaptiveSmooth
MitchellOneQuarter
MitchellOneSixth
MitchellNegOneSixth
MitchellNegOneThird
Lanczos4
Lanczos5
WithoutFlagsMask
Flag_WrapX
Flag_WrapY
Flag_AlphaWeighted

◆ ProcessLinearPixelsAction

enum class FImageCore::ProcessLinearPixelsAction

strong

Enumerator
ReadOnly
Modified

Function Documentation

◆ ComputeChannelLinearMinMax()

void FImageCore::ComputeChannelLinearMinMax	(	const FImageView &	InImage,
		FLinearColor &	OutMin,
		FLinearColor &	OutMax
	)

Compute the min/max of each channel to get value ranges Colors are converted to float Linear Gamma

Parameters

InImage	- The image to scan
OutMin	- filled with the minimum of the color channels
OutMax	- filled with the maximum of the color channels

Design:

FImage pixels are dense, so we just treat them as a bunch of samples of U8 or U16. (no need to look at width/height/slices at all).

We ignore the pixel format other than knowing it is a U8 or U16 channel. We always work on 16-byte pieces, which can be varying number of pixels.

Cut the data into pieces for parallel processing (16-byte aligned). Find the 16-byte vector min/max on those pieces. Then gather min/max of the 16-bytes from each piece.

For the tail portion that may not be a full 16 bytes, we can just replicate a pixel to fill 16 bytes and use the same 16-byte aligned min/max routines, so no special tail case is required.

Finally once we have the 16-byte min/max of the whole image, we run the Generic fallback which does the horizontal min/max inside that vector, and also handles correctly interpreting whether it is 16xG8 or 4xBGRA8 or whatever.

◆ ComputeImageLinearAverage()

IMAGECORE_API FLinearColor FImageCore::ComputeImageLinearAverage ( const FImageView & Image )

ComputeImageLinearAverage compute the average linear color of the image image can be any pixel format parallel processing is used, but the result is not machine-dependent

◆ CopyImage()

IMAGECORE_API void FImageCore::CopyImage	(	const FImageView &	SrcImage,
		const FImageView &	DestImage
	)

Copy FImageView data, converting pixels if necessary. Sizes must match. Dest must already be allocated. For a function that allocates Dest, use CopyTo().

Because FImage implicitly convers to FImageView, these functions can be used on FImage types as well.

Parameters

SrcImage	- The source image to copy from.
DestImage	- The destination image to copy to. (the FImageView is const but what it points at is not)

Copies an image accounting for format differences. Sizes must match.

Parameters

SrcImage	- The source image to copy from.
DestImage	- The destination image to copy to.

◆ CopyImageRGBABGRA()

IMAGECORE_API void FImageCore::CopyImageRGBABGRA	(	const FImageView &	SrcImage,
		const FImageView &	DestImage
	)

Copy Image, swapping RB, SrcImage must be BGRA8 or RGBA16. Sizes must match. Dest must already be allocated. Src == Dest is okay for in-place transpose.

Parameters

SrcImage	- The source image to copy from.
DestImage	- The destination image to copy to. (the FImageView is const but what it points at is not)

◆ CopyImageTo2U16()

IMAGECORE_API void FImageCore::CopyImageTo2U16	(	const FImageView &	SrcImage,
		const FImageView &	DestImage
	)

Copy Image to 2xU16 ; Dest Image should be allocated as BGRA8 but is actually 2xU16 Sizes must match. Dest must already be allocated.

Parameters

SrcImage	- The source image to copy from.
DestImage	- The destination image to copy to. (the FImageView is const but what it points at is not)

◆ DetectAlphaChannel()

bool FImageCore::DetectAlphaChannel ( const FImageView & InImage )

Detects whether or not the image contains an alpha channel where at least one texel is != 255.

Parameters

InImage - The image to look for alpha in.

Returns: True if the image both supports and uses the alpha channel. False otherwise.

◆ ENUM_CLASS_FLAGS()

FImageCore::ENUM_CLASS_FLAGS ( EResizeImageFilter )

◆ ImageParallelFor()

template<typename Lambda >

void FImageCore::ImageParallelFor	(	const TCHAR *	DebugName,
		const FImageView &	Image,
		const Lambda &	Func
	)

inline

◆ ImageParallelForComputeNumJobs()

IMAGECORE_API int32 FImageCore::ImageParallelForComputeNumJobs	(	const FImageView &	Image,
		int64 *	pRowsPerJob
	)

FImage is tight packed in memory with slices adjacent to each other so we can just treat it was a 2d image with Height *= NumSlices

eg. make "ImagePart" of 16384 pixels, and make "Rows" for the FLinearColor pass that are always exactly 512 pixels

◆ ImageParallelForComputeNumRows()

int64 FImageCore::ImageParallelForComputeNumRows ( const FImageView & Image )

inline

◆ ImageParallelForGetOneRowView()

FImageView FImageCore::ImageParallelForGetOneRowView	(	const FImageView &	Image,
		int64	Y
	)

inline

◆ ImageParallelForMakePart()

IMAGECORE_API int64 FImageCore::ImageParallelForMakePart	(	FImageView *	Part,
		const FImageView &	Whole,
		int64	JobIndex,
		int64	RowsPerJob
	)

◆ ImageParallelProcessLinearPixels()

template<typename Lambda >

void FImageCore::ImageParallelProcessLinearPixels	(	const TCHAR *	DebugName,
		const FImageView &	Image,
		const Lambda &	Func
	)

inline

◆ ProcessLinearPixels()

template<typename Lambda >

void FImageCore::ProcessLinearPixels	(	const FImageView &	Image,
		const Lambda &	Func,
		int64	StartY
	)

inline

ProcessLinearPixels act on image pixels as TArrayView64<FLinearColor> Colors pass a lambda that is : ProcessLinearPixelsAction ProcessPixels(TArrayView64<FLinearColor> Colors) return ProcessLinearPixelsAction::Modified if you modify colors return ProcessLinearPixelsAction::ReadOnly if you only read and don't change colors your lambda will be called on one row at a time

◆ ResizeImage()

IMAGECORE_API void FImageCore::ResizeImage	(	const FImageView &	SourceImage,
		const FImageView &	DestImage,
		EResizeImageFilter	Filter = `EResizeImageFilter::Default`
	)

◆ ResizeImageAllocDest() [1/2]

IMAGECORE_API void FImageCore::ResizeImageAllocDest	(	const FImageView &	SourceImage,
		FImage &	DestImage,
		int32	DestSizeX,
		int32	DestSizeY,
		ERawImageFormat::Type	DestFormat,
		EGammaSpace	DestGammaSpace,
		EResizeImageFilter	Filter = `EResizeImageFilter::Default`
	)

◆ ResizeImageAllocDest() [2/2]

IMAGECORE_API void FImageCore::ResizeImageAllocDest	(	const FImageView &	SourceImage,
		FImage &	DestImage,
		int32	DestSizeX,
		int32	DestSizeY,
		EResizeImageFilter	Filter = `EResizeImageFilter::Default`
	)

◆ ResizeImageInPlace() [1/2]

IMAGECORE_API void FImageCore::ResizeImageInPlace	(	FImage &	Image,
		int32	DestSizeX,
		int32	DestSizeY,
		ERawImageFormat::Type	DestFormat,
		EGammaSpace	DestGammaSpace,
		EResizeImageFilter	Filter = `EResizeImageFilter::Default`
	)

◆ ResizeImageInPlace() [2/2]

IMAGECORE_API void FImageCore::ResizeImageInPlace	(	FImage &	Image,
		int32	DestSizeX,
		int32	DestSizeY,
		EResizeImageFilter	Filter = `EResizeImageFilter::Default`
	)

◆ ResizeTo()

void FImageCore::ResizeTo	(	const FImageView &	SourceImage,
		FImage &	DestImage,
		int32	DestSizeX,
		int32	DestSizeY,
		ERawImageFormat::Type	DestFormat,
		EGammaSpace	DestGammaSpace
	)

Copies and resizes the image to a destination image with the specified size and format. Resize is done using bilinear filtering; gamma correct (light linear) by converting through RGBA32F.

legacy API, prefer ResizeImage

Parameters

DestImage	- The destination image.
DestSizeX	- Width of the resized image
DestSizeY	- Height of the resized image
DestFormat	- The destination image format.
DestSRGB	- Whether the destination image is in SRGB format.

◆ SanitizeFloat16AndSetAlphaOpaqueForBC6H()

void FImageCore::SanitizeFloat16AndSetAlphaOpaqueForBC6H ( const FImageView & InOutImage )

Clamp Float16 colors which aren't encodable in the BC6H format RGB is clamped to non-negative and finite A is set to 1.0

Parameters

InOutImage is modified. Must be RGBA16F. (the FImageView is const but what it points at is not)

◆ ScaleChannelsSoMinMaxIsInZeroToOne()

bool FImageCore::ScaleChannelsSoMinMaxIsInZeroToOne ( const FImageView & ImageToModify )

If the image has any values outside the [0,1] range, rescale that edge of the domain so that it is in [0,1] does not affect images that were previously in [0,1]

also does not change the side of the domain that is not out of bounds eg. values in [0.25,200.0] will be rescaled to [0.25,1.0]

returns bool if any change was made

If the input format is U8 or U16, no change will ever be made and this will return false.

This can be useful if you want to save an HDR/float image to a U8 image format for visualization. This is equivalent to what's called the "UNorm" transformation by the RenderTarget ReadPixels functions.

◆ SetAlphaOpaque()

void FImageCore::SetAlphaOpaque ( const FImageView & InImage )

Change alpha channel to opaque, if present

Parameters

InImage - The image to look for alpha in.

◆ TransformToWorkingColorSpace()

void FImageCore::TransformToWorkingColorSpace	(	const FImageView &	InLinearImage,
		const FVector2d &	SourceRedChromaticity,
		const FVector2d &	SourceGreenChromaticity,
		const FVector2d &	SourceBlueChromaticity,
		const FVector2d &	SourceWhiteChromaticity,
		UE::Color::EChromaticAdaptationMethod	Method,
		double	EqualityTolerance = `1.e-7`
	)

Apply a color space transformation from the source chromaticities to the engine's working color space.

Parameters

InLinearImage	- The image to convert, which must be linear.
SourceRedChromaticity	- The red chromaticity coordinate of the source color space.
SourceGreenChromaticity	- The green chromaticity coordinate of the source color space.
SourceBlueChromaticity	- The blue chromaticity coordinate of the source color space.
SourceWhiteChromaticity	- The white chromaticity coordinate of the source color space.
Method	- The chromatic adapation method.
EqualityTolerance	- The tolerance for the source and working color space chromaticities to be considered equal, bypassing the transform.

◆ TransposeImageRGBABGRA()

IMAGECORE_API void FImageCore::TransposeImageRGBABGRA ( const FImageView & Image )

Swap RB channels on Image. Image must be BGRA8 or RGBA16.

Parameters

Image is modified (the FImageView is const but what it points at is not)

Classes

Enumerations

Functions

Detailed Description

Enumeration Type Documentation

◆ EResizeImageFilter

◆ ProcessLinearPixelsAction

Function Documentation

◆ ComputeChannelLinearMinMax()

◆ ComputeImageLinearAverage()

◆ CopyImage()

◆ CopyImageRGBABGRA()

◆ CopyImageTo2U16()

◆ DetectAlphaChannel()

◆ ENUM_CLASS_FLAGS()

◆ ImageParallelFor()

◆ ImageParallelForComputeNumJobs()

◆ ImageParallelForComputeNumRows()

◆ ImageParallelForGetOneRowView()

◆ ImageParallelForMakePart()

◆ ImageParallelProcessLinearPixels()

◆ ProcessLinearPixels()

◆ ResizeImage()

◆ ResizeImageAllocDest() [1/2]

◆ ResizeImageAllocDest() [2/2]

◆ ResizeImageInPlace() [1/2]

◆ ResizeImageInPlace() [2/2]

◆ ResizeTo()

◆ SanitizeFloat16AndSetAlphaOpaqueForBC6H()

◆ ScaleChannelsSoMinMaxIsInZeroToOne()

◆ SetAlphaOpaque()

◆ TransformToWorkingColorSpace()

◆ TransposeImageRGBABGRA()