Going Beyond Bitrate: Adaptation Based on Video Context
summary
Claims that adapting bitrate per scene saves bandwidth without perceptual loss by extracting encoder-derived metrics (PSNR, SSIM, VMAF) per chunk and shipping them as quality metadata with segments. Demonstrates replacing high-bitrate segments in low-complexity scenes (examples: cartoons vs action) with 1 Mbps instead of 3 Mbps while keeping visual quality, and using saved bandwidth to sustain higher quality on complex scenes. Implements per-title encoding ideas on the client as per-scene adaptation.
AI-written summary · about
chapters (4)
related talks
17mDelivering better video encodes for legacy devices
15mTowards Measuring Perceptual Video Quality & Why
22mProviding better video experiences for the next billion users.
26mBandwidth Ballet: Optimizing Video Performance for Web Maestros
11mFrom Artifacts to Art: A Data-Driven Guide to Android Video Encoding
19mRethinking the Video Transcoding Pipeline: How Pre-Filtering Holds the Key to Efficiency and Quality
13mSENSEI: Aligning Video Streaming Quality with Dynamic User Sensitivity
17mContent Aware Encoding for low latency live streaming encoders using deep learning