Quantification

The signal is now sampled. The next stage in digitization is quantization. The bit rate is directly linked to this.

To quantize a signal, we need to define a quantization law and scale. The quantization scale has been set to 256 levels, i.e. 8 bits, and the law has been chosen to be linear. The choice of 8-bit words may seem obvious, given the components available and the habit of working on 8-bit, but it needs to be justified. For black and white, 7 bits of quantization suffice; for color – to which the eye is less sensitive – 6 bits of quantization suffice, hence the choice to work in bytes. This satisfies both luminance and chrominance constraints.

What's more, a better signal-to-noise ratio is obtained by choosing a logarithmic compression law, and the choice of this linear law may seem anomalous. But the video signal is already γ-corrected....