A lightweight neural pre-encoder for unmodified libx264 achieves -27.62% mean BD-VMAF on UVG and similar gains on MCL-JCV by applying bounded pixel adjustments learned via a hybrid differentiable proxy.
MCL-JCV: A JND-based H.264/AVC video quality assessment dataset
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.IV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Kelvin v1.0: A Neural Pre-Encoder for H.264: A standards-compliant learned preprocessor with -27.62% BD-VMAF on UVG
A lightweight neural pre-encoder for unmodified libx264 achieves -27.62% mean BD-VMAF on UVG and similar gains on MCL-JCV by applying bounded pixel adjustments learned via a hybrid differentiable proxy.