Colocated with ICME2018 July 23rd – 27th, 2018, San Diego, USA
The significant growth of 3D sensing technologies, along with the increased interest in adopting 3D representations in imaging systems, are shaping the way media content is created and used. In this environment, point clouds have become a viable solution, since they provide a practical way for capture, storage, delivery and rendering in augmented reality, mixed reality, virtual reality, medical imaging and 3D printing applications, among others. A point cloud can be defined as a collection of three-dimensional points in space representing the external surface of an object. Each sample is defined by its position, which is obtained by the measured or reconstructed X, Y, and Z coordinates. Associated features can also be used in conjunction with the coordinate data to provide further information, such as the point’s color, intensity, normal, size, curvature, and specularity.
Business and technology specialists predict that immersive applications will become mainstream by 2020. Point clouds and mesh representations are expected to be modalities that will be extensively exploited, since they allow users to visualize static or dynamic contents in a natural and immersive way. Although polygon mesh has been the prevailing form in computer graphics for 3D representations, point cloud is expected to be the future solution in real-time applications due to several important features: (a) High resolution point clouds can be directly acquired nowadays using professional or even consumer devices (e.g., Microsoft Kinect, Intel RealSense, LIDAR). In future, conventional cameras are also expected to produce point cloud content either by making use of use depth sensors or by processing multiple images captured from different positions. (b) Polygon meshes require sophisticated reconstruction techniques when updating the content for every additional collection of points, with significant cost in terms of time and computational complexity. Using point clouds, the captured objects could be updated directly by rendering the additional points. (c) Compared to the mesh-based compression algorithms, point clouds do not require storage, delivery and decoding of connectivity information (i.e., mesh topology and faces). The conversion to mesh representation for coding or rendering can also be avoided in many situations. Despite all advantages mentioned above, point cloud content of high quality requires a very large volume of information which can hinder their use in both professional and consumer applications. There is a need for an interchange and delivery format that can efficiently reduce the volume of information required for minimal or no impact on their quality. Despite efforts in the scientific and standardization, an efficient compression algorithm for point clouds does not exist yet. This challenge solicits contributions that demonstrate efficient compression of point cloud data. Moreover, new evaluation methodologies are sought. Furthermore, additional publicly accessible point cloud content along with evidence for compression efficiency as well as other attractive features are also accepted.
Data sets / APIs / URL:
The data set, the subjective evaluation methodology used for quality assessment, the software used for visualization and processing of point cloud data, the software for objective assessment of point cloud data used in the grand challenge and the anchors are provided in the url below:
The same URL will also contain additional software tools for different objective quality metrics and additional or alternative anchors against point cloud coding techniques submitted by proponents will be compared to.
1. Evaluation of Compression Efficiency
Each point cloud content and any associated data used for the reconstruction and interaction with it shall be compressed at several bit rates on a selected subset from the above-mentioned database. The performance will be evaluated using both subjective evaluations and objective metrics in an interactive approach. All subjective evaluations methodologies and objective quality metrics will be carried out by two independent test laboratories (EPFL and UBI) and crossed checked to validate their reliability and repeatability.
2. Subjective evaluation methodologies and objective metrics
Proposals for new evaluation methodologies and objective metrics are welcome along with proposals for point cloud coding to demonstrate any one or several of the following:
- Alternative subjective evaluation methodologies that can assess subjective quality of point clouds after compression, processing or manipulation.
- Alternative objective quality metrics that are well correlated with subjective assessment of point cloud quality over a range of manipulations including but not limited to compression.
- Methods that assess the quality of interaction with a compressed point cloud such as assessment of random access, resilience to errors, progressive decoding, free-navigation, etc. either in conventional display environments or in immersive environment making use of Head Mounted Display (VR, AR, MR).
Proposals of such evaluation methodologies and objective metrics should provide a very detailed description along with an executable and a source code of their submission.
3. Additional content
Proposals for new content motivated by existing or emerging use cases are welcome along with proposals for point cloud coding and evaluation methodologies. Such content should be made freely available for research and standardization purposes.
Grand Challenge proponents’ submission deadline:
March 19, 2018 extended to March 26, 2018
Grand Challenge proponents’ optional paper submission deadline:
March 19, 2018 extended to March 26, 2018
Grand Challenge acceptance notification: April 23, 2018
Grand Challenge optional camera-ready paper submission deadline: May 11, 2018
Submissions of point cloud coding algorithms along with optional evaluation methodologies and any additional content should provide a detailed technical description. In addition, proponents can decide to submit a paper following the same guidelines in ICME 2018 that will be peer reviewed in a similar way. In all cases, proponents are required to submit material to validate the performance of their submission according to the procedure outlined in “ICME2018-Grand Challenge.pdf” available in the above mentioned URL. Submissions for evaluation methodologies and objective metrics or additional content should be made along with a point cloud algorithm. The optional paper could be included as part of the ICME 2018 proceedings and published on IEEE Xplore if accepted after peer review.
Touradj Ebrahimi (EPFL)
Antonio Pinheiro (Instituto de Telecomunicacoes and UBI)
Anthony Vetro (Mitsubishi Electric Research Labs)
Touradj Ebrahimi [email protected]