Media Compression NUS SOC CS 5248 2019 Roger

Media Compression NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

You are Here Encoder Decoder Middlebox Receiver Sender Network NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Demo n Sinéad O'Connor (Nothing Compares 2 U) NUS. SOC. SWS 3021 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Why compress? n “Bandwidth Not Enough” n “Disk Space Not Enough” n Size of Uncompressed DVD Movie = NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Why compress? n “Bandwidth Not Enough” n “Disk Space Not Enough” n Size of Uncompressed DVD Movie = (720 x 576) pixels x 3 bytes x 25 fps x 60 sec/min x 120 min = 208. 6 GB n NTSC: 29. 97 fps (30/1. 001); PAL 25 fps NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Optical Disc Formats (1) n CD: ~650 MB n Video. CD: codec MPEG-1 n 1 X max. read speed: 1. 5 Mb/s n DVD: n 4. 7 (4. 38) GB (single layer) n 8. 5 (7. 92) GB (dual layer) n Single and dual sided (up to 18 GB) n 1 X max. read speed: ~10 Mb/s n Video codec: MPEG-2 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Optical Disc Formats (2) n Blu-ray n Capacity: 25 GB and 50 GB n 1 X speed: 36 Mb/s n Video codec: VC-1, H. 264, MPEG-2 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

JPEG Compression NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Original Image (1153 KB) 1: 1

Original Image (1153 KB) 3. 5: 1

Original Image (1153 KB) 17: 1

Original Image (1153 KB) 27: 1

Original Image (1153 KB) 72: 1

Original Image (1153 KB) 192: 1

Compression Ratio Quality Size Ratio Raw TIFF 1153 KB 1: 1 Zipped TIFF 982 KB 1. 2: 1 Q=100 331 KB 3. 5: 1 Q=70 67 KB 17: 1 Q=40 43 KB 27: 1 Q=10 16 KB 72: 1 Q=1 6 KB 192: 1 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Magic of JPEG n Throw away information we cannot see, i. e. , based on human visual system: n Color information n “High frequency signals” n Rearrange data for good compression n Use standard compression algorithms NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Discard color information Y V NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang) U

Color Sub-sampling n The subsampling scheme is commonly expressed as a n n three part ratio (e. g. 4: 2: 2). The parts are (in their respective order): Luma (Y) horizontal sampling reference (originally, as a multiple of 3. 579 MHz in the NTSC television system). Cr (U) horizontal factor (relative to first digit). Cb (V) horizontal factor (relative to first digit), except when zero. Zero indicates that Cb horizontal factor is equal to second digit, and, in addition, both Cr and Cb are subsampled 2: 1 vertically. Zero is chosen for the bandwidth calculation formula to remain correct. To calculate required bandwidth factor relative to 4: 4: 4, one needs to sum all the factors and divide the result by 12. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Color Sub-sampling 4: 4: 4 4: 2: 0 4: 2: 2 4: 1: 1 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

4: 2: 2 Sub-sampling Y V NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang) U

Original Image (1153 KB) 4: 2: 0

Original Image (1153 KB) “ 4: 1: 0”

Discrete Cosine Transform Demo NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Quantization DC 242 65 23 5 8 8 8 -54 -10 -4 -2 8 8 8 16 13 6 3 5 8 8 16 32 2 1 -1 -2 / 8 8 16 32 64 Quantization Table NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang) 30 8 = 2 0 -6 -1 0 0 0 0 0 AC

Differential Coding 30 8 2 0 25 3 1 0 27 3 1 0 6 -1 0 0 2 1 0 0 0 4 0 1 0 0 0 0 30 8 2 0 -5 3 1 0 2 3 1 0 6 -1 0 0 2 1 0 0 0 4 0 1 0 0 0 0 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Zig-zag ordering 27 3 1 0 2 1 0 0 4 0 1 0 0 0 27, 3, 2, 4, 1, 1, 0, 0, 0, 1, 0, 0 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Run-Length Encoding 27 3 1 0 2 1 0 0 4 0 1 0 0 0 27, 3, 2, 4, 1, 1, 0, 0, 0, 1, 0, 0 (27, 1) (3, 1) (2, 1), (4, 1), (1, 2), (0, 5), (1, 1), (0, 4) NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Idea: Motion JPEG n Compress every frame in a video as JPEG n DVD-quality video = 208. 6 GB n Reduction ratio = 27: 1 n Final size = 7. 7 GB NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Video Compression NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Original Frame 1 By (c) copyright 2006, Blender Foundation / Netherlands Media Art Institute / www. elephantsdream. org - Screenshot from "Elephants Dream" http: //orange. blender. org/download, CC BY 2. 5, https: //commons. wikimedia. org/w/index. php? curid=7395123 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Difference Between 2 Frames By (c) copyright 2006, Blender Foundation / Netherlands Media Art Institute / www. elephantsdream. org - Screenshot from "Elephants Dream" http: //orange. blender. org/download, CC BY 2. 5, https: //commons. wikimedia. org/w/index. php? curid=7395129 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Motion Compensated Difference By (c) copyright 2006, Blender Foundation / Netherlands Media Art Institute / www. elephantsdream. org - Screenshot from "Elephants Dream" http: //orange. blender. org/download, CC BY 2. 5, https: //commons. wikimedia. org/w/index. php? curid=7395132 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Temporal Redundancy NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Motion Estimation NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Bi-directional Prediction NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Motion Vectors NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

H. 261 P-Frame I-Frame NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

MPEG-1 B-Frame NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

MPEG Frame Pattern (1) n HDV GOP example NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

MPEG Frame Pattern (2) n Example display sequence: n IBBPBBP … n Example encoding sequence: n IPBBPBB NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Compression Ratio Frame Typical Ratio I 10: 1 P 20: 1 B 50: 1 NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Sequence sequence header: • width • height • frame rate • bit rate • : NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

GOP: Group of Picture gop header: • time • : NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Picture pic header: • number • type (I, P, B) • : NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Picture NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Slice NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Slice n Slices are important in the handling of errors. If the bitstream contains an error, the decoder can skip to the start of the next slice. n Having more slices in the bitstream allows better error concealment, but uses bits that could otherwise be used to improve picture quality (worse compression). NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Macroblock NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Block 1 Macroblock = NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang) Y Y U Y Y V

Structure Summary NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

For I-Frame n Every macroblock is encoded independently (“I-macroblock”) NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

For P-Frame n Every macroblock is either n I-macroblock n a motion vector + error terms with respect to a previous I/P-frame (“Pmacroblock”) NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

For B-Frame n Every macroblock is either n I-macroblock n P-macroblock n a motion vector + error terms wrt a future I/P-frame n 2 motion vectors + error terms wrt a previous/future I/P-frame NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

MPEG-1/2 File Formats n (Packetized) Elementary streams, ES & PES n Program streams PS (reliable mediums, e. g. , DVD) n Transport streams TS (for lossy mediums, e. g. , on-air broadcast) Video Source MPEG-2 Elementary Encoder Packetizer MPEG encoded streams Audio Source MPEG-2 Elementary Encoder Data Source NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang) PES: *. m 2 v PES: *. m 2 a Packetizer Systems Layer MUX TS: *. ts *. m 2 t *. mpg Transport Stream Packetizer Flow chart © Manish Karir

Review: MPEG structure n ES, PS, TS: elementary stream, program stream, transport stream n Sequence n GOP: group of pictures n Picture Important: n Slice Codec standards are n Macroblock essentially defined by their bitstreams! n Block NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Container Formats n Two common container formats: n n MPEG-2 TS (Transport Stream) ISO Base Media File Format (ISOBMFF) n Container formats encapsulate the actual media streams and can contain various actual media types. n E. g. , TS is still used today with H. 264 Container (ISOBMFF) Media data, e. g. , video and audio (H. 264) NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

MPEG Decoding (I-Frame) 101000101 Entropy Decoding Dequantize IDCT NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

MPEG Decoding (P-Frame) 101000101 Entropy Decoding Dequantize IDCT Prev Frame NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang) +

MPEG Decoding (B-Frame) 101000101 Entropy Decoding Future Frame Dequantize IDCT AVG Prev Frame NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang) +

There is much more … n Half-pel motion prediction n Skipped macroblock n Different sizes of macroblocks n Motion vectors across multiple frames n etc. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Codecs in Daily Life (1) MPEG Standards Bit-rate Usage MPEG-1 1. 5 Mbps VCD MPEG-2 3 -45 Mbps DVD, SVCD, HDTV MPEG-4/ H. 264/AVC Scalable, ½ MPEG-2 H. 265/HEVC Scalable, ½ H. 264 Quick. Time, Div. X, AVCHD, Cable TV, You. Tube, … New generation, 4 K content “H. 266” Scalable, ½ H. 265 Next generation, 8 K content NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Codecs in Daily Life (2) © The Register, 16 Sep 2015, Nigel Whitfield NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Codec Comparison n “M-JPEG” (e. g. , DV) versus “MPEG” Compression Technique “M-JPEG” (I-frames only) “MPEG” (Temporal compression) Compression ratio Low (10: 1 to 30: 1) High (>100: 1) Editing (frame-accurate) Easy Difficult Encoding/decoding complexity Symmetric Asymmetric Processing latency Low to Medium High Multi-generation loss Medium High n No “perfect” codec -> application dependent NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

High-Definition n Standard by ATSC n 18 different sub-formats n 720 p and 1080 i are the most interesting n 1280 x 720 x 60 p, 1920 x 1080 x 60 i (30 p) n 1080 p is non-standard, but available n 1. 4 Gb/s raw bandwidth n 10 – 20 Mb/s compressed (distribution, broadcast) n 100 – 135 Mb/s compressed (pro tapes: DVCPROHD, HDCAM; for editing) NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Consumer HD n HDV: MPEG-2 n 19 (720 p) / 25 Mb/s (1080 i) n Tape format n http: //www. hdv-info. org n AVCHD: H. 264 n 5 to 25 Mb/s n Hard disk format n http: //www. avchd-info. org/ NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Current Popular Codec: H. 264 n “Same quality at half the rate” n Encoding complexity: ~4 X n How: n Variable block size motion compensation n Multiple reference frames n Deblocking filter, … n Also called MPEG-4 Part 10 or AVC or MPEG-4/AVC NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Open Codec: VP 8 n Google bought On 2 Technologies in 2010, which developed VP 8. n Open-source license. (H. 264 needs to be licensed for use) n Similar coding efficiency and quality as H. 264. n Uses the Web. M file format. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Current Codec: H. 265/HEVC n High Efficiency Video Coding (HEVC). n “Same quality at half the rate” (over H. 264/MPEG-4 AVC). n Very high encoding complexity. n Supports progressive scanned frame rates and display resolutions from QVGA (320 x 240) up to 1080 p (1920 x 1080) and Ultra HDTV (7680 x 4320). NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Current Codec: H. 265/HEVC n At WWDC 2017, Apple introduced two new camera formats that are included in i. OS 11: HEVC and HEIF. n Starting with i. Phone 6, HEVC was the format for Face. Time. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Open Codecs: VP 9, AV 1 n VP 9 is an open and royalty free video coding format developed by Google. n Successor to VP 8. n Supported by most web browsers (except Safari). n Used by You. Tube. n Successor: AV 1 (2019). NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

H. 265/HEVC Primary Changes n Expansion of the pattern comparison and n n difference-coding areas from 16× 16 pixel to sizes up to 64× 64. Improved variable-block-size segmentation, improved "intra" prediction within the same picture. Improved motion vector prediction and motion region merging. Improved motion compensation filtering. An additional filtering step called sample-adaptive offset filtering. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Deblocking Filter n With heavy compression, blocking artifacts become visible. Deblocking filters are used in both H. 264 and H. 265. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Next Gen Codec: H. 266/FVC n Future Video Codec (FVC), expected in 2021. n Further improvements to 4 K and enable “even 8 K, 12 K and 16 K”. n Compression ratio target >50% over HEVC. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Hands-On n Download source code, compile and play with ffmpeg n mpeg_stat n Video ‘Surfing_short. m 2 t’ from course web site (98 MB, HDV, transport stream) n n Try different MPEG-1/2 encoding parameter NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Impact on Systems Design How to package data into packets? How to deal with packet loss? How to deal with bursty traffic? How to predict decoding time? : : n n NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)

Summary n Compression removes data for which the human visual system (HVS) is not sensitive. n Current codecs are mostly based on DCT and motion compensation. n Container formats (MPEG-2 TS, ISOBMFF) are important for system designers. n Codec standards are essentially defined by their bitstreams. NUS. SOC. CS 5248 -2019 Roger Zimmermann (based in part on slides by Ooi Wei Tsang)