lti::splitImageToYUV Class Reference
[Color Analysis]

Computes the YUV values from a given RGB color representation (rgbPixel). More...

#include <ltiSplitImageToYUV.h>

Inheritance diagram for lti::splitImageToYUV:

[legend]

Collaboration diagram for lti::splitImageToYUV:

[legend]

List of all members.

Public Member Functions
	splitImageToYUV ()
	splitImageToYUV (const splitImageToYUV &other)
virtual	~splitImageToYUV ()
virtual const char *	getTypeName () const
splitImageToYUV &	copy (const splitImageToYUV &other)
splitImageToYUV &	operator= (const splitImageToYUV &other)
virtual functor *	clone () const
virtual bool	apply (const rgbPixel &pixel, float &c1, float &c2, float &c3) const
virtual bool	apply (const rgbPixel &pixel, ubyte &c1, ubyte &c2, ubyte &c3) const
virtual bool	apply (const image &img, channel &c1, channel &c2, channel &c3) const
virtual bool	apply (const image &img, channel8 &c1, channel8 &c2, channel8 &c3) const

Detailed Description

Computes the YUV values from a given RGB color representation (rgbPixel).

In the literature, technical and scientific, there is a confusion between the color spaces YUV, YCrCb and YPbPr. Poynton in http://www.poynton.com/notes/colour_and_gamma/ColorFAQ.html explains that YUV is usually never correctly meant, because the color space normally used for component digital video is the YCrCb (ITU-RS601 or CCIR-601). Other devices use the YPbPr, but the "real" YUV is usually not employed.

The LTI-Lib provides all three spaces:

YCrCb: lti::splitImageToYCbCr used by IEEE 1394 FireWire cameras
YPbPr: lti::splitImageToYPbPr used by some WebCams
YUV: lti::splitImageToYUV did they really meant to use this?

This functor splits an RGB images into the color space YUV as described in the Poynton reference given above.

$\begin{bmatrix} Y \\ U \\ V \end{bmatrix} = \begin{bmatrix} 0.299 & 0.587 & 0.114 \\ -0.147141189 & -0.288869157 & 0.436010346 \\ 0.614975383 & -0.514965121 & -0.100010262 \end{bmatrix} \begin{bmatrix} R \\ G \\ B \end{bmatrix}$

Here, RGB are values in the intervall [0,1]. Within this range, the excursion for Y is also from 0.0 to 1.0, the excursion of U is from -0.436 to 0.436 and V varies from -0.615 to 0.615. These ranges are respected in the implentation for float types (lti::channel).

If you use ubyte (lti::channel8), then the values for U and V are linearly mapped from 0 to 1, which results in the mapping

 Y =  0.299*R + 0.587*G + 0.114*B 
 U = -0.169*R - 0.331*G + 0.500*B + 128.0 
 V =  0.500*R - 0.419*G - 0.081*B + 128.0

The former mapping is equivalent to the color space YPbPr scaled in such a way that "black" is mapped to Y=0 and "white" to Y=255, and the Pb and Pr channels, which acquire values from -0.5 to 0.5, are linearly mapped to the inteval [0,255] as well. In other words, for ubyte operations it is equivalent to use this functor or lti::splitImageToYPbPr (which is maybe better named as the value ratios are at least kept there).

Constructor & Destructor Documentation

lti::splitImageToYUV::splitImageToYUV ( )

default constructor

lti::splitImageToYUV::splitImageToYUV ( const splitImageToYUV & other )

copy constructor

Parameters:

other

the object to be copied

virtual lti::splitImageToYUV::~splitImageToYUV ( ) [virtual]

destructor

Member Function Documentation

virtual bool lti::splitImageToYUV::apply	(	const image &	img,
		channel8 &	c1,
		channel8 &	c2,
		channel8 &	c3
	)			const `[virtual]`

Split image into 8-bit channels.

N.B.: when casting the transformation result to unsigned shorts (8-bit channel) major rounding errors will occur.

As a result, the merging operation might produce negative values or values > 1, which are truncated subsequently. When accurate Y, U and V values are required, prefer float values.

Implements lti::splitImage.