2.2 Digital video

The process of digitising analogue video involves the three basic operations of filtering, sampling and quantisation. The filtering operation is employed to avoid the aliasing artefacts of the follow-up sampling process. The filtering applied to the luminance can be different to that for chrominance, owing to different bandwidth requirements.

Filtered luminance and chrominance signals are sampled to generate a discrete time signal. The minimum rate at which each component can be sampled is its Nyquist rate and corresponds to twice the signal bandwidth. For a PAL system this is in the range of 10–11 MHz. However, due to the requirement to make the sampling frequency a harmonic of the analogue signal line frequency, the sampling rate for broadcast quality signals has been recommended by CCIR to be 13.5 MHz, under recommendation CCIR-601 [2]. This is close to three times the PAL subcarrier frequency. The chrominance sampling frequency has also been defined to be half the luminance sampling frequency. Finally, sampled signals are quantised to eight-bit resolution, suitable for video broadcasting applications.

It should be noted that colour space recommended by CCIR-601 is very close to the PAL system. The precise luminance (Y) and chrominance (C_b and C_r) equations under this recommendation are:

(2.2)

The slight departure from the PAL parameters is due to the requirement that, in the digital range, Y should take values in the range of 16–235 quantum levels. Also, the normally AC chrominance components of U and V are centred on the grey level 128, and the range is defined from 16 to 240. The reasons for these modifications are:

to reduce the granular noise of all three signals in later stages of processing
to make chrominance values positive to ease processing operations (e.g. storage).

Note that despite the unique definition for Y, C_b and C_r, the CCIR-601 standard for European broadcasting is different from that for North America and the Far East. In the former, the number of lines per frame is 625 and the number of frames per second is 25. In the latter these values are 525 and 30, respectively. The number of samples per active line, called picture elements (pixels) is 720 for both systems. In the 625-line system, the total number of pixels per line, including the horizontal blanking, is 13.5 MHz times 64 μs, equal to 864 pixels. Note also that despite the differences in the number of lines and frames rates, the number of pixels generated per second under both CCIR-601/625 and CCIR-601/525 is the same. This is because in digital television we are interested in the active parts of the picture, and the number of active television lines per frame in CCIR-601/625 is 576 and the total number of pixels per second becomes equal to 720 × 576 × 25 = 10 368 000. In CCIR-601/525 the number of active lines is 480, and the total number of pixels per second is 720 × 480 × 30 = 10 368 000.

The total bit rate is then calculated by considering that there are half the luminance pixels for each of the chrominance pixels, and with eight bits per pixel, the total bit rate becomes 10 368 000 × 2 × 8 = 165 888 000 bits/s. Had we included all the horizontal and vertical blanking, then the total bandwidth would be 13.5 × 10⁶ × 2 × 8 = 216 Mbit/s. Either of these values is much greater than the equivalent analogue bandwidth, hence the video compression to reduce the digital bit rate is very demanding. In the following chapters we will show how such a huge bit rate can be compressed down to less than 10 Mbit/s, without noticeable effect on picture quality.