Abstract:A segmentation map, either static or dynamic, refers to a two-dimensional picture that may vary with time and indicates the segmentation label per pixel. Both the semantic map and the occupancy map in video-based point cloud compression (V-PCC) belong to the segmentation map we referred to. The semantic map can work for many machine vision tasks like tracking and has been used as a layer of image representation in some image compression methods. The occupancy map constitutes a part of the point cloud coding bitstream. Since segmentation maps are widely used, how to efficiently compress them is of interest. We propose a segmentation map lossless compression scheme namely CC-SMC, exploiting the nature of segmentation maps that usually contain limited colors and sharp edges. Specifically, we design a chain coding-based scheme combined with quadtree-based block partitioning. For intraframe coding, one block is partitioned recursively with a quadtree structure, until the block contains only one color, is smaller than a threshold, or satisfies the defined chain coding condition. We revise the three-orthogonal chain coding method to incorporate contextual information and design effective intraframe prediction methods. For interframe coding, one block may find a reference block; the chain difference between the current and the reference blocks is coded. We implement the proposed scheme and test it on several different kinds of segmentation maps. Compared with advanced lossless image compression techniques, our proposed scheme obtains more than 10% bits reduction as well as more than 20% decoding time-saving. The code is available at https://github.com/Yang-Runyu/CC-SMC.

Unsigned Manhattan chain code

An Efficient Multi-Resolution Chain Coding

A universal chain code compression method

Chain code based on independent edge number

A Modified Vertex-Based Shape Coding Algorithm

On the Chain Code of a Line.

NU-MCC: Multiview Compressive Coding with Neighborhood Decoder and Repulsive UDF

Chain Code-Based Occupancy Map Coding for Video-Based Point Cloud Compression

An Efficient Contour Tracking and Coding Algorithm Based on Vertex Chain Code

Max Markov Chain

CC-SMC: Chain Coding-Based Segmentation Map Lossless Compression

GenUDC: High Quality 3D Mesh Generation with Unsigned Dual Contouring Representation

Study on Chain Code Methods for Triangular Grids

Coding for the unsourced B-channel with erasures: enhancing the linked loop code

A Non-Manhattan Hexagon/Triangle Placement Paradigm

Euler Characteristic Computation by Means of a Chain Code Applied to Binary Images

A Versatile Point Cloud Compressor Using Universal Multiscale Conditional Coding - Part II: Attribute

Multilevel Constructions: Coding, Packing and Geometric Uniformity

Coding for the unsourced A-channel with erasures: the linked loop code

BCH Based U-UV Codes and Its SCL Decoding