Figure 3.16 The early days of color television. A digital image can be represented using threecolor values per pixel.” However, the colorimetric storing of images in RGB space is inefficient, as there is consid- erable correlation between the channels. Alternatively, when images or video are to be compressed, they are usually converted into YCbCr color space. In ¥CbCr two chroma (Cb, Cr) components The Ima channel contains “brightness” information; it is essentially a gray scale version of the image. The chroma vahues are color offsets, which show how much a pixel deviates from grey scale in the blue (Cb) and red (Cr) directions. For most natural images, the RGB to YCbCr con- version strongly de-correlates the color channels, so the Y, Cb, and Cr components can be coded independently. In YCbCr space, the bandwidth of an image tends to be concentrated in the Y channel. This leaves the Cand Cr channels with less information, so they can be represented with fewer bits. Another advantage of the YCbCr space comes from the characteristics of the human visual system. Humans are more sensitive to brightness information than color information. Consequently, the chrom signals (Cb, Cr) can be downsampled relative to the Juma (Y) without significant loss of perceived image quality. Chroma downsampling is almost always done when compressing image or video data. In YCbCr4:4:4 format, the chroma values (Cb, Cr) are sampled at the same rate as Iuma (Y). This for- mat is rarely used, except in professional applications (see Figure 3.17), In YCbCr 4:2:2 format, the chroma values (Cb, Cr) are downsampled by a factor of two relative to the Juma (Y) in the horizontal direction. Most higher end digital video formats use YCbCr 4:2:2 sampling (see Figure 3.18) Figure 3.18 Conversion from By Figure 3.19 Conversions from Bayer Cell toVCbCr 4:2:0. in compressed images and video is YCbCr 4:2:0 (see Figure 3.19) Tn YCbCr 4:2:0 format, the chroma signals are downsampled by a factor of two in both the hori zontal and vertieal directions. It should be noted 1 there are different chroma positions used in YCbCr 4:2:0 format. Sometimes the chroma samples are considered to be halfway between the luma samples vertically, or in the center of a group of four luma samples. In the downsampling process the chroma channels should be low pass filtered to limit aliasing In Figure 3.20, we can see our original RGB example scene and, in Figure 3.2, its YCbCr waveform scope representation. The luminance ¥ (on theleft, Figure3.21) is used to multiply the © components (chroma blue, and chroma red, ht) to reproduce the subset of the original color data, irom a small Figure 3.20 ICAS “Serious News!igure 3.21 YCBCr waveform monitor screen grab of the scer Y'CbCr color encoding is used to separate out a Juma signal (Y') that can be stored with high resolution or transmitted at high bandwidth, and two chroma components (Cb and Cr) that can be bandwidih- reduced, subsampled, and compressed for improved system efficiency, Figure Figure 3.25 The red chroma component The blue and red chroma components are effec tively half the bandwidth of the green luma com- ponent, not a problem in many ordinary viewing circumstances, but a huge potential source of head- aches for visual effects work Figure 3.26 A typical green screen composite before and afer. Blue-sereen or green-sereen mattes are more dif ficult to extract from chroma-subsampled images, as the difference in sampling resolution between color records in composites can cause fringing and matte edge lines that are clearly visible and nearly impos- sible to mitigate. Big screen? VFX? Shoot 4:44, or