BitRIC: Efficient Neural Compression of LiDAR Range Images via Hierarchical Bitplanes

The rapid evolution of 3D scene understanding relies heav-ily on high-resolution LiDAR, which generates massive data volumesthat impose significant bottlenecks on transmission and storage. WhileRange Image Compression (RIC) provides a more structured and com-putationally efficient alternative to conventional Point Cloud Compres-sion (PCC), existing RIC frameworks frequently fail to achieve an op-timal trade-off between coding efficiency and real-time performance. Tobridge this gap, we propose BitRIC, a novel learning-based frameworktailored for high-performance LiDAR range image compression. BitRICdecomposes raw range images into hierarchical bitplanes and processesthem through a coarse-to-fine autoregressive entropy model. At its core, ashared contextual backbone leverages cross-bitplane dependencies to in-form two specialized pathways: a Probability Estimation Module (PEM)that predicts probability maps for lossless entropy coding, and a One-Step Refinement Module (ORM) that reconstructs residual details dur-ing bitplane truncation for lossy compression. Extensive experiments onthe Waymo Open Dataset and SemanticKITTI demonstrate that BitRICachieves superior rate-distortion performance, yielding over 60% BD-BRsavings compared to standard image codecs, while offering higher down-stream task accuracy. Furthermore, BitRIC supports real-time lossy andnear-real-time lossless processing, making it a highly practical candidatefor real-world applications.