Table 4.

Summary of 2D-based inter-frame compression methods.

Category	Method	Type	Main Features	Setup	Dataset	Performance	Source Code
Traditional	RI-LZW (1984) [80]	Lossy	Applies the LZW codec on a sequence of range images created from LiDAR	Intel Core i5-4210U	Velodyne HDL-64 sensor	PSNR: 63 [39]	Open source
	RI-MJ2 (2003) [60]	Lossy	Applies the MJ2 codec on a sequence of range images created from LiDAR	Intel Core i5-4210U	Velodyne HDL-64 sensor	PSNR: 63 [39]	Open source
	RI-H.264 (2014) [78]	Lossless	Applies the H.264 codec on a sequence of range images created from LiDAR	Intel Core i7-4770	Velodyne HDL-64 sensor	bpp: 2.41	Open source
	RI-LayerJPEG (2016) [81]	Lossy	Applies the JPEG codec to layered range images created from LiDAR	Not disclosed	Velodyne HDL-64 sensor	PSNR: 49–80	Not disclosed
	RT-ST (2020) [85]	Lossless	Uses iterative plane fitting to exploits both spatial and temporal redundancies	Intel Core i5-7500 and Nvidia mobile TX2	SemanticKITTI	CR: 40–90	Not disclosed
	PC-SLAM (2021) [82,83]	Lossy	Uses location and orientation information for LiDAR data compression	Intel Core i7-7820X	Velodyne HDL-64 sensor	bpp: 3.61–6.68	Not disclosed
	CLUSTER-ICP (2021) [84]	Lossless/Lossy	Uses CLUSTER [67], registration-based inter-prediction and lossless compression on residuals	i5-6300HQ 2.3 GHz w/ 4GB RAM	KITTI	CR: 9.47–41.49	Not disclosed
	FLiCR (2022) [79]	Lossy	Uses H.264 video codec on lossy RI for edge-assisted online perception	Nvidia Jetson AGX Xavier	KITTI	CR: 21.26–215.85	Not disclosed
Learning	RT-S-PCC-U-NET (2019) [86]	Lossless	Uses U-Net [87] to reduce temporal redundancies in a sequence of frames	Intel Core i7-7820X w/ Nvidia GeForce GTX 1080	Velodyne HDL-64 sensor	bpp: 2–4.5	Not disclosed
	Inter-Inserting (2022) [91]	Lossless	Uses plane fitting on RangeNet++ [72] RI’s segments and an interpolation-based network for temporal redundancy removal	Desktop w/ Nvidia TITAN RTX	KITTI	CR: 14.56–32.36	Not disclosed
	CLUSTER-LSTM (2022) [89]	Lossless/Lossy	Uses CLUSTER [67] for intra-prediction and convolutional LSTM cells for inter-frame compression	Intel 2.2GHz i7 w/ Nvidia GPU and 16GB RAM	KITTI	CR: 24.39–63.29	Not disclosed
	RIDDLE (2022) [92]	Lossy	Uses a deep model to predict the next pixel values based on current and past LiDAR scans and delta encoding to compress the data	Nvidia Tesla V100	Waymo Open and KITTI	bpp: 3.65–4.3	Not disclosed
	BPNet RAFC (2022) [90]	Lossy	Uses a frame prediction network to inter-frame prediction and floating-point lossy encoder for I- and B-frame residuals	Intel Core i7-7700K w/ Nvidia GTX 1080Ti and 16GB RAM	KITTI	bpp: 5.7–7.3	Not disclosed
	BIRD-PCC (2023) [88]	Lossless	Uses R-PCC [75] as intra-frame compression and U-Net [87] w/ a binary mask for inter-frame compression	Not disclosed	SemanticKITTI and KITTI-360	bpp: 1.7–4.2	Not disclosed