Multimedia Compression Standards: JPEG, MPEG, H.26x and Audio
What this domain tests
Matching each standard to its structure, application and terminology. Items ask about the JPEG pipeline, which MPEG part does what, how I/P/B frames and GOPs work, or which audio layer a file extension implies.
Core rules and concepts
- JPEG. Level-shift by −128, apply an 8×8 DCT, divide by a quality-scaled quantisation table, reorder coefficients in zigzag order, then entropy-code: the DC coefficient is DPCM-coded against the previous block, AC coefficients use run-length plus Huffman (baseline) or arithmetic coding. Modes: baseline sequential, extended, progressive, and lossless (predictive, no DCT). JPEG 2000 instead uses wavelet transforms, EBCOT coding, region-of-interest and scalability.
- MPEG-1 targets VCD (about 352×288 at 1.5 Mbit/s) and defines Audio Layers I, II and III; MP3 is Layer III. MPEG-2 targets DVD and broadcast, adds interlaced tools and transport streams. MPEG-4 adds object-based coding and BIFS; its Part 10 is H.264/AVC. MPEG-7 is a metadata description standard, MPEG-21 a framework/DRM standard - neither compresses video.
- H.26x. H.261 for ISDN, H.263 for low bit rate, H.264/AVC (integer 4×4 transform, CABAC, multiple reference frames, in-loop deblocking), H.265/HEVC (coding tree units up to 64×64 with quadtree splitting).
- Frame types. I frames are independently coded, P frames predict from earlier frames, B frames use past and future references, hence bidirectional prediction and reordering.
- Audio. Psychoacoustic models drive bit allocation; MP3 uses a 32-band filterbank plus MDCT; AAC improves on it; AC-3 is the Dolby broadcast standard.
Worked example
A GOP is displayed as I B B P. Because each B frame references the following P frame, the encoder must transmit I, then P, then the two B frames. A decoder therefore buffers the P frame, reconstructs the B frames, and outputs in display order. If the question asks which frame can be decoded after losing an earlier packet, the answer is the next I frame, because P and B depend on their references.
Traps that cost marks
- Describing MPEG-7 or MPEG-21 as compression standards.
- Thinking baseline JPEG uses wavelets - that is JPEG 2000.
- Believing B frames are always better; they raise latency and require buffering.
- Confusing the file extension with the algorithm: MP3 is MPEG-1 Layer III, not "MPEG-3".