Image Class Reference

represents one image (actually a SDL_Context wrapper ..) More...

#include <Image.h>

List of all members.

Public Types
enum	{ GRAY_MASK = 1 }
	Converts the image to vector. More...
Public Member Functions
	Image (unsigned xs, unsigned ys)
	new image of undefined content in default RGBA pixel format
void	resize (unsigned xs, unsigned ys)
	~Image ()
void	toFile (const std::string &fileName)
	Save image to file (BMP).
Image *	subImage (unsigned width, unsigned height, unsigned xOffset=0, unsigned yOffset=0) const
	Create a sub-image of the original.
void	setSubImage (const Image &src, unsigned xo=0, unsigned yo=0)
	Copy contents of image from the src image to given offset.
Color	getAverageColor () const throw (Exception)
Image *	toRGBA () const
	Image pixel format will be converted to RGBA (one pixel = one byte) Unless you know the input image format, you should convert image with this function before making training examples from a loaded image.
Image *	clone () const
	copy this image (the result is independent)
Image *	upsideDown () const
	flip image
Image *	toGray () const
	Image will be converted to 32-RGBA (one pixel = four bytes).
unsigned	getWidth () const
unsigned	getHeight () const
unsigned	getBPP () const
Vector	toVector (uint8_t bytes_mask) const
void	display (unsigned xo=0, unsigned yo=0) const
	display to the video window (offsets are relative to win coords) if !gl, the video surface is flipped automatically (so far) - under GL you have to do it yourself (in GL mode, the offset is passed directly to glRasterPos --> it's up to caller to give it sense
void	clear ()
void	rect (const Color &c, uint x1, uint y1, uint x2, uint y2)
void	putPixel (const Color &c, uint x, uint y)
	draw a pixel TODO: avoid frequent locking???
Color	getPixel (uint x, uint y) const
	get a pixel
real	getPixelGray (uint x, uint y) const
	return the pixel as [0,1] gray
Static Public Member Functions
Image *	fromFile (const std::string &fileName)
Image *	fromVector (const Vector &v, unsigned width, bool gray)
	Creates an image from a vector of [0,1] reals.
Protected Member Functions
	Image (SDL_Surface *i)
	Create image wrapping the given surface.
int	glFormat () const
void	setGlPixelStore (bool unpack) const
SDL_Surface *	createSameSurface (uint width, uint height) const
	create void surface with the same flags as this
SDL_Surface *	createSameSurface () const
void	glDraw () const
	direct GL draw (assuming rasterPosition is set)
void	colorless ()
	Average the color components.
uint32_t	color2SDL (const Color &c) const
uint32_t	getPixelRaw (uint x, uint y) const
Color	sdl2color (uint32_t p) const
Protected Attributes
SDL_Surface *	surface

---

Detailed Description

represents one image (actually a SDL_Context wrapper ..)

Unless you force it by to* functions, you can assume nothing about pixel format of the image! Typical use would therefore be:

   Image *raw = Image::fromFile(...); if(!raw) ...
   Image *rgb_img = raw->toRGBA();
   delete(raw);
   if(!rgb_img) ...
   rgb_img-toVector(...)
   delete(rgb_img);
   

NOTE: some const functions may in theory change state of the Image because they lock the surface Image uses window coordinates where 0,0 is top-left

Todo:

use exceptions instead of NULL ?

---

Member Enumeration Documentation

anonymous enum

Converts the image to vector. Each choosen byte is mapped into [0,1] real (0=black) Parameters: bytes_mask.  Sets the bytes you want from each pixel (system-dependent-endian) - e.g. 0x2 | 0x8 means you want 0x00FF0000 and 0x000000FF XXX Note: you have to make sure the image is in proper pixel format - this function just blindly reads given bytes Enumeration values: GRAY_MASK

---

Constructor & Destructor Documentation

Image (  unsigned  xs, unsigned  ys )

new image of undefined content in default RGBA pixel format

~Image (   )

Image (  SDL_Surface *  i  )  [inline, protected]

Create image wrapping the given surface. The created image owns the surface and will free in destructor.

---

Member Function Documentation

void clear (   )

Image* clone (   )  const

copy this image (the result is independent)

uint32_t color2SDL (  const Color &  c  )  const [protected]

void colorless (   )  [protected]

Average the color components. The surface must have 4-bytes-per-pixel. The result is a grayscale-looking surface (but it is still RGBA!)

SDL_Surface* createSameSurface (   )  const [protected]

SDL_Surface* createSameSurface (  uint  width, uint  height )  const [protected]

create void surface with the same flags as this

void display (  unsigned  xo = 0, unsigned  yo = 0 )  const

display to the video window (offsets are relative to win coords) if !gl, the video surface is flipped automatically (so far) - under GL you have to do it yourself (in GL mode, the offset is passed directly to glRasterPos --> it's up to caller to give it sense

Image* fromFile (  const std::string &  fileName  )  [static]

Returns: NULL on failure

Image* fromVector (  const Vector &  v, unsigned  width, bool  gray )  [static]

Creates an image from a vector of [0,1] reals. The real images are mapped into color component bytes (R,G,B or Y,Y,Y (graY)) Parameters: width  x-size of the created image gray  if true, only one value will be used for the three color components (--> resulting in grayscale) else each pixel has three components

Color getAverageColor (   )  const throw (Exception)

Parameters: out  - [3] vector for color components - will be [0,1]

unsigned getBPP (   )  const

unsigned getHeight (   )  const

Color getPixel (  uint  x, uint  y )  const

get a pixel

real getPixelGray (  uint  x, uint  y )  const

return the pixel as [0,1] gray

uint32_t getPixelRaw (  uint  x, uint  y )  const [protected]

unsigned getWidth (   )  const

void glDraw (   )  const [protected]

direct GL draw (assuming rasterPosition is set)

int glFormat (   )  const [protected]

void putPixel (  const Color &  c, uint  x, uint  y )

draw a pixel TODO: avoid frequent locking???

void rect (  const Color &  c, uint  x1, uint  y1, uint  x2, uint  y2 )

void resize (  unsigned  xs, unsigned  ys )

Color sdl2color (  uint32_t  p  )  const [protected]

void setGlPixelStore (  bool  unpack  )  const [protected]

void setSubImage (  const Image &  src, unsigned  xo = 0, unsigned  yo = 0 )

Copy contents of image from the src image to given offset. The source image can be smaller, but not larger

Image* subImage (  unsigned  width, unsigned  height, unsigned  xOffset = 0, unsigned  yOffset = 0 )  const

Create a sub-image of the original. Resulting image has the same pixel format. Returns: NULL on failure

void toFile (  const std::string &  fileName  )

Save image to file (BMP).

Image* toGray (   )  const

Image will be converted to 32-RGBA (one pixel = four bytes). But all three RGB components are equal, so the image is grayscale-only. TODO: could we use one-byte SDL surface? We'd probably need to use palleted surface so the displayed image would be correct.

Image* toRGBA (   )  const

Image pixel format will be converted to RGBA (one pixel = one byte) Unless you know the input image format, you should convert image with this function before making training examples from a loaded image.

Vector toVector (  uint8_t  bytes_mask  )  const

Image* upsideDown (   )  const

flip image Returns: the flipped clone

---

Member Data Documentation

SDL_Surface* surface [protected]

---

The documentation for this class was generated from the following file:

• Image.h

---