08.01.2013 Views

DigitalVideoAndHDTVAlgorithmsAndInterfaces.pdf

DigitalVideoAndHDTVAlgorithmsAndInterfaces.pdf

DigitalVideoAndHDTVAlgorithmsAndInterfaces.pdf

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

To understand the mathematical<br />

details of color transforms,<br />

described in this section, you should<br />

be familiar with linear (matrix)<br />

algebra. If you are unfamiliar with<br />

linear algebra, see Strang, Gilbert,<br />

Introduction to Linear Algebra,<br />

Second Edition (Boston: Wellesley-Cambridge,<br />

1998).<br />

Table 22.6 Example primaries<br />

are used to explain the necessity<br />

of signal processing in<br />

accurate color reproduction.<br />

Eq 22.1 This matrix is based upon<br />

R, G, and B components with<br />

unusual spectral distributions. For<br />

typical R, G, and B, see Eq 22.8.<br />

Eq 22.2<br />

because light power cannot go negative. So we cannot<br />

build a real display that responds directly to XYZ. But as<br />

you will see, the concept of negative SPDs – and<br />

nonphysical SPDs or nonrealizable primaries – is very<br />

useful in theory and in practice.<br />

There are many ways to choose nonphysical primary<br />

SPDs that correspond to the X(λ), Y(λ), and Z(λ) colormatching<br />

functions. One way is to arbitrarily choose<br />

three display primaries whose power is concentrated at<br />

three discrete wavelengths. Consider three display<br />

SPDs, each of which has some amount of power at<br />

600 nm, 550 nm, and 470 nm. Sample the X(λ), Y(λ),<br />

and Z(λ) functions of the matrix given earlier in Calculation<br />

of tristimulus values by matrix multiplication, on<br />

page 218, at those three wavelengths. This yields the<br />

tristimulus values shown in Table 22.6:<br />

Red, 600 nm Green, 550 nm Blue, 470 nm<br />

X 1.0622 0.4334 0.1954<br />

Y 0.6310 0.9950 0.0910<br />

Z 0.0008 0.0087 1.2876<br />

These coefficients can be expressed as a matrix, where<br />

the column vectors give the XYZ tristimulus values<br />

corresponding to pure red, green, and blue at the<br />

display, that is, [1, 0, 0], [0, 1, 0], and [0, 0, 1]. It is<br />

conventional to apply a scale factor in such a matrix to<br />

cause the middle row to sum to unity, since we wish to<br />

achieve only relative matches, not absolute:<br />

⎡X<br />

⎤ ⎡0.<br />

618637 0. 252417 0. 113803⎤<br />

⎡R<br />

⎤<br />

⎢ ⎥ ⎢<br />

⎥ ⎢ 600nm⎥<br />

⎢Y<br />

⎥ = ⎢0.<br />

367501 0. 579499 0. 052999⎥<br />

• ⎢G550nm⎥<br />

⎢<br />

⎣<br />

Z ⎥ ⎢<br />

⎥ ⎢ ⎥<br />

⎦ ⎣<br />

0. 000466 0. 005067 0. 749913<br />

⎦ B<br />

⎣⎢<br />

470nm⎦⎥<br />

That matrix gives the transformation from RGB to XYZ.<br />

We are interested in the inverse transform, from XYZ to<br />

RGB, so invert the matrix:<br />

⎡R<br />

⎤<br />

600nm ⎡ 2. 179151 −0. 946884 −0.<br />

263777⎤<br />

⎡X<br />

⎤<br />

⎢ ⎥ ⎢<br />

⎥ ⎢ ⎥<br />

⎢G550nm⎥<br />

= ⎢−1.<br />

382685 2. 327499 0. 045336⎥<br />

• ⎢Y<br />

⎥<br />

⎢ ⎥<br />

B ⎢<br />

Z<br />

⎣⎢<br />

470nm⎦⎥<br />

⎣<br />

0. 007989 −0.<br />

015138 1. 333346⎥<br />

⎢ ⎥<br />

⎦ ⎣ ⎦<br />

CHAPTER 22 COLOR SCIENCE FOR VIDEO 241

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!