Learn how does video CODEC works under the hood
summary
Covers how codecs actually cut 250 GB of raw 1080p video down (chroma subsampling halves color data to ~125 GB, then transforms and entropy coding bring it to ~1.3 GB and playable at ~6 Mbps). Explains DCT transform, block partitioning, zig‑zag + entropy coding (run‑length/Huffman), and the difference between cheap frame‑diffs and motion estimation (saves more bits). Shows an FFmpeg demo and argues HEVC/H.265 improves compression by larger blocks, more prediction directions and CPU tradeoffs. Good if you want mechanisms and numbers.
AI-written summary · about
chapters (12)
- 0:00Introduction and overview
- 1:40What is video compression needed
- 6:57Digital video basics
- 10:08Human vision and color models
- 13:05Chroma subsampling explained
- 17:45Temporal compression techniques
- 20:14Spatial compression and transforms
- 23:19How DCT compression works
- 26:56Encoding and bitrate reduction
- 30:34Decoder workflow and summary
- 32:23Codec efficiency and trade-offs
- 34:34Q&A and closing
related talks
28mThe Genesis of a Universal Video Transcoder
15mFast Lossless Depth Image Compression
22mMovie Gen Video: State-of-the-art Video Generation
24mBeyond the Frame: Using Computer Vision to Unlock the Potential of Video Data
37mRendering performance inside out
31mBuilding Camera Features in a Snap
11mAIRCODE: Hidden Screen-Camera Communication on an Invisible and Inaudible Dual Channel
10mRollingEvidence: Autoregressive Video Evidence via Rolling Shutter Effect