Digital Media Representation
What this domain tests
How continuous sound, images and video become finite numbers, and what those numbers cost in bits. Questions ask you to compute bit rates and storage, interpret colour-depth or sampling notation, or explain an artefact such as aliasing or banding.
Core rules and concepts
- Sampling and quantisation. The Nyquist–Shannon theorem requires a sampling rate above twice the highest frequency; 44.1 kHz for audio follows from a ~20 kHz hearing limit. Too low a rate produces aliasing, which must be prevented by an anti-aliasing low-pass filter before sampling, not removed afterwards. Quantisation error sets the noise floor: SQNR ≈ 6.02n + 1.76 dB for n-bit uniform quantisation, so 16-bit audio gives roughly 98 dB.
- Image representation. An image is a pixel matrix with spatial resolution, colour depth and a colour model. Indexed (palette) colour stores an index into a lookup table; true colour stores RGB components directly. An alpha channel carries per-pixel opacity.
- Colour models. RGB is additive and device-oriented (displays); CMYK is subtractive (print); HSV/HSB is perceptual and used in pickers; YUV/YCbCr separates luma from chroma, which enables chroma subsampling because human vision resolves colour coarsely.
- Chroma subsampling notation J:a:b describes a 4×2 reference block. 4:4:4 is unfiltered; 4:2:2 halves horizontal chroma; 4:2:0 halves chroma both horizontally and vertically, cutting chroma data to one quarter.
- Video. Frame rate, interlaced fields versus progressive frames, pixel aspect ratio versus display aspect ratio, and colour primaries/gamma all affect both quality and bit cost.
Worked example
Uncompressed 1920×1080, 24-bit, 25 fps video: 1920 × 1080 × 3 bytes = 6,220,800 bytes per frame; × 25 = 155,520,000 bytes/s ≈ 148 MiB/s ≈ 1.24 Gbit/s. Audio CD: 44,100 × 2 bytes × 2 channels = 176,400 bytes/s ≈ 1.41 Mbit/s. The comparison explains why video dominates storage and why compression is mandatory rather than optional: the audio stream is under 0.2% of the raw video rate.
Traps that cost marks
- Mixing decimal and binary units: 1 Mbit/s = 10^6 bit/s, but 1 MiB = 2^20 bytes.
- Assuming 4:2:0 halves the total data; it reduces chroma to a quarter, so total reduction depends on the luma/chroma split.
- Believing dpi changes the pixel count; it is a print-density metadata value.
- Thinking a higher sampling rate fixes aliasing that already occurred during capture.