Image I/O

Loading and saving image files from disk. More...

Functions
bool	loadImageRGBA (const char *filename, float4 **cpu, float4 **gpu, int *width, int *height, const float4 &mean_pixel=make_float4(0, 0, 0, 0))
	Load a color image from disk into CUDA memory with alpha, in float4 RGBA format with pixel values 0-255. More...
bool	saveImageRGBA (const char *filename, float4 *cpu, int width, int height, float max_pixel=255.0f, int quality=100)
	Save a float4 RGBA image to disk. More...
bool	loadImageRGB (const char *filename, float3 **cpu, float3 **gpu, int *width, int *height, const float3 &mean_pixel=make_float3(0, 0, 0))
	Load a color image from disk into CUDA memory, in float3 RGB format with pixel values 0-255. More...
bool	loadImageBGR (const char *filename, float3 **cpu, float3 **gpu, int *width, int *height, const float3 &mean=make_float3(0, 0, 0))
	Load a color image from disk into CUDA memory, in float3 BGR format with pixel values 0-255. More...

Detailed Description

Loading and saving image files from disk.

Function Documentation

loadImageBGR()

bool loadImageBGR	(	const char *	filename,
		float3 **	cpu,
		float3 **	gpu,
		int *	width,
		int *	height,
		const float3 &	mean = make_float3(0, 0, 0)
	)

Load a color image from disk into CUDA memory, in float3 BGR format with pixel values 0-255.

Supported image file formats by loadImageBGR() include:

• JPEG
• PNG
• TGA
• BMP
• GIF
• PSD
• HDR
• PIC
• PNM (PPM/PGM binary)

This function loads the image into shared CPU/GPU memory, using the functions from cudaMappedMemory.h

Parameters

[in]	filename	Path to the image file to load from disk.
[out]	cpu	Reference to pointer that will be set to the CPU buffer containing the image that will be allocated. This buffer will be allocated by loadImageRGBA() in CUDA mapped memory, so it is shared between CPU/GPU. cpu will be the pointer to this shared buffer in the CPU's address space. There is physically one buffer in memory.
[out]	gpu	Reference to pointer that will be set to the GPU buffer containing the image that will be allocated. This buffer will be allocated by loadImageRGBA() in CUDA mapped memory, so it is shared between CPU/GPU. gpu will be the pointer to this shared buffer in the GPU's address space. There is physically one buffer in memory.
[in,out]	width	Pointer to int variable that gets set to the width of the image in pixels. If the width variable contains a non-zero value when it's passed in, the image is resized to this desired width. Otherwise if the value of width is 0, the image will be loaded with it's dimensions from the file on disk.
[in,out]	height	Pointer to int variable that gets set to the height of the image in pixels. If the height variable contains a non-zero value when it's passed in, the image is resized to this desired height. Otherwise if the value of height is 0, the image will be loaded with it's dimensions from the file on disk.
[in]	mean_pixel	Mean pixel subtraction is applied using this specified color (mean pixel subtraction is typically performed before processing an image with neural networks). By default the value is zero, so no mean pixel subtraction is done.

loadImageRGB()

bool loadImageRGB	(	const char *	filename,
		float3 **	cpu,
		float3 **	gpu,
		int *	width,
		int *	height,
		const float3 &	mean_pixel = make_float3(0, 0, 0)
	)

Load a color image from disk into CUDA memory, in float3 RGB format with pixel values 0-255.

Supported image file formats by loadImageRGB include:

• JPEG
• PNG
• TGA
• BMP
• GIF
• PSD
• HDR
• PIC
• PNM (PPM/PGM binary)

This function loads the image into shared CPU/GPU memory, using the functions from cudaMappedMemory.h

Parameters

[in]	filename	Path to the image file to load from disk.
[out]	cpu	Reference to pointer that will be set to the CPU buffer containing the image that will be allocated. This buffer will be allocated by loadImageRGBA() in CUDA mapped memory, so it is shared between CPU/GPU. cpu will be the pointer to this shared buffer in the CPU's address space. There is physically one buffer in memory.
[out]	gpu	Reference to pointer that will be set to the GPU buffer containing the image that will be allocated. This buffer will be allocated by loadImageRGBA() in CUDA mapped memory, so it is shared between CPU/GPU. gpu will be the pointer to this shared buffer in the GPU's address space. There is physically one buffer in memory.
[in,out]	width	Pointer to int variable that gets set to the width of the image in pixels. If the width variable contains a non-zero value when it's passed in, the image is resized to this desired width. Otherwise if the value of width is 0, the image will be loaded with it's dimensions from the file on disk.
[in,out]	height	Pointer to int variable that gets set to the height of the image in pixels. If the height variable contains a non-zero value when it's passed in, the image is resized to this desired height. Otherwise if the value of height is 0, the image will be loaded with it's dimensions from the file on disk.
[in]	mean_pixel	Mean pixel subtraction is applied using this specified color (mean pixel subtraction is typically performed before processing an image with neural networks). By default the value is zero, so no mean pixel subtraction is done.

loadImageRGBA()

bool loadImageRGBA	(	const char *	filename,
		float4 **	cpu,
		float4 **	gpu,
		int *	width,
		int *	height,
		const float4 &	mean_pixel = make_float4(0, 0, 0, 0)
	)

Load a color image from disk into CUDA memory with alpha, in float4 RGBA format with pixel values 0-255.

Supported image file formats by loadImageRGBA() include:

• JPEG
• PNG
• TGA
• BMP
• GIF
• PSD
• HDR
• PIC
• PNM (PPM/PGM binary)

This function loads the image into shared CPU/GPU memory, using the functions from cudaMappedMemory.h

Parameters

[in]	filename	Path to the image file to load from disk.
[out]	cpu	Reference to pointer that will be set to the CPU buffer containing the image that will be allocated. This buffer will be allocated by loadImageRGBA() in CUDA mapped memory, so it is shared between CPU/GPU. cpu will be the pointer to this shared buffer in the CPU's address space. There is physically one buffer in memory.
[out]	gpu	Reference to pointer that will be set to the GPU buffer containing the image that will be allocated. This buffer will be allocated by loadImageRGBA() in CUDA mapped memory, so it is shared between CPU/GPU. gpu will be the pointer to this shared buffer in the GPU's address space. There is physically one buffer in memory.
[in,out]	width	Pointer to int variable that gets set to the width of the image in pixels. If the width variable contains a non-zero value when it's passed in, the image is resized to this desired width. Otherwise if the value of width is 0, the image will be loaded with it's dimensions from the file on disk.
[in,out]	height	Pointer to int variable that gets set to the height of the image in pixels. If the height variable contains a non-zero value when it's passed in, the image is resized to this desired height. Otherwise if the value of height is 0, the image will be loaded with it's dimensions from the file on disk.
[in]	mean_pixel	Mean pixel subtraction is applied using this specified color (mean pixel subtraction is typically performed before processing an image with neural networks). By default the value is zero, so no mean pixel subtraction is done.

saveImageRGBA()

bool saveImageRGBA	(	const char *	filename,
		float4 *	cpu,
		int	width,
		int	height,
		float	max_pixel = 255.0f,
		int	quality = 100
	)

Save a float4 RGBA image to disk.

Supported image file formats by saveImageRGBA() include:

• JPEG
• PNG
• TGA
• BMP
• HDR

Parameters

filename	Desired path of the image file to save to disk.
cpu	Pointer to the buffer containing the image in CPU address space.
width	Width of the image in pixels.
height	Height of the image in pixels.
max_pixel	The maximum pixel value of this image, by default it's 255 for images in the range of 0-255. If your image is in the range of 0-1, pass 1.0 as this value. Then the pixel values of the image will be rescaled appropriately to be stored on disk (which expects a range of 0-255).
quality	Indicates the compression quality level (between 1 and 100) to be applied for JPEG and PNG images. A level of 1 correponds to reduced quality and maximum compression. A level of 100 corresponds to maximum quality and reduced compression. By default a level of 100 is used for maximum quality and reduced compression. Note that this quality parameter only applies to JPEG and PNG, other formats will ignore it.