Neural Network Assisted Depth Map Packing for Compression Using Standard Hardware Video Codecs

TL;DR

This paper presents a neural network-assisted method for packing depth maps into RGB frames to enable efficient compression with standard video codecs, optimizing precision to balance quality and bitrate.

Contribution

It introduces a variable precision packing scheme guided by neural network predictions, improving depth map compression without requiring specialized hardware extensions.

Findings

Neural network accurately predicts optimal packing precision for depth maps.

Proposed method achieves near-optimal compression quality at low overhead.

Integration into game engines is feasible with modern hardware.

Abstract

Depth maps are needed by various graphics rendering and processing operations. Depth map streaming is often necessary when such operations are performed in a distributed system and it requires in most cases fast performing compression, which is why video codecs are often used. Hardware implementations of standard video codecs enable relatively high resolution and framerate combinations, even on resource constrained devices, but unfortunately those implementations do not currently support RGB+depth extensions. However, they can be used for depth compression by first packing the depth maps into RGB or YUV frames. We investigate depth map compression using a combination of depth map packing followed by encoding with a standard video codec. We show that the precision at which depth maps are packed has a large and nontrivial impact on the resulting error caused by the combination of the packing scheme and lossy compression when bitrate is constrained. Consequently, we propose a variable precision packing scheme assisted by a neural network model that predicts the optimal precision for each depth map given a bitrate constraint. We demonstrate that the model yields near optimal predictions and that it can be integrated into a game engine with very low overhead using modern hardware.