AI Paper Atlas

Research map / Video / Video segmentation and tracking

On the choice of Perception Loss Function for Learned Video Compression

Sadaf Salehkalaibar, Truong Buu Phan, Jun Chen, Wei Yu, Ashish J Khisti

NeurIPS 2023 · Video › Video segmentation and tracking · approach: Learned video compression · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use PLF-JD instead of PLF-FMDMovingMNIST causal video compression · MovingMNIST, R1=12 bits, R2=12MSE distortion: 1.43±0.02 -> 2.26 ± 0.03Hurt
Use PLF-JD instead of PLF-FMDMovingMNIST causal video compression · MovingMNIST, R1=12 bits, R2=∞MSE distortion: 0.0 -> 2.18 ± 0.02Hurt
Use PLF-JD instead of PLF-FMDMovingMNIST causal video compression · MovingMNIST, R1=∞ bits, R2=1MSE distortion: 1.74 ± 0.02 -> 2.05 ± 0.03Hurt
Use PLF-JD instead of PLF-FMDMovingMNIST causal video compression · MovingMNIST, R1=∞ bits, R2=3.17MSE distortion: 0.76 ± 0.01 -> 0.79 ± 0.01No clear effect
Use PLF-JD to preserve the joint frame distributionCausal video decoder · MovingMNIST, low-rate GOP of 3 framestemporal consistencyImproved
Use PLF-JD rather than PLF-FMDCausal video decoder · MovingMNIST and UVG, low-rate initial frameerror correction across framesHurt
Train the encoder for MSE and choose the perception objective at the decoderVideo compression representations · MovingMNIST, rate-distortion-perception tradeoffdistortion-perception tradeoffNo clear effect
Convert an MMSE reconstruction to a reconstruction satisfying zero-PLF-FMDFixed video encoder · General sourcesMSE distortion: at most a factor of 2 increaseHurt
Use zero-PLF-JD for a general fixed encoder rather than MSE aloneFixed video encoder · General sourcesMSE distortionHurt
Adapt an MMSE representation using decoder-side linear transforms and possibly additive noiseGauss-Markov video compression · Gauss-Markov source, any point in the iDP rate regionrate-distortion-perception tradeoffImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy