Plenoptics for Virtual Reality

Publications

Quality Evaluation of Point Cloud Coding Solutions

Publication . Prazeres, João Pedro Casanova; Pinheiro, António Manuel Gonçalves

Point clouds experienced a large increase in popularity. From gaming to medical applications, autonomous driving, and urban mapping, point clouds have been widely used in the current technological world. As the demand for point cloud content increases, the need for efficient point cloud coding solutions also increases. Access to such solutions is important for efficient storage and transmission of point cloud data, because they are typically represented by huge amounts of information. It is, however, crucial to have access to quality methods that accurately benchmark point cloud coding solutions. This allows the developers of such solutions to accurately test their codec in several different environments, adjusting the codec development accordingly. In the past, subjective quality models were established to assess the quality of images and videos. Based on this knowledge, new models were developed for point cloud content, though there are crucial differences due to the 3D nature of point clouds. Recently, the growing popularity of learning-based codecs led to a new analysis of the performance of the developed quality models, as the caused distortions tend to be different from those created by the traditional coding technologies. This thesis aims to research those well established subjective quality models in order to assess the performance of point cloud coding solutions, namely the ones that are learningbased. Furthermore, it was also important to understand how the current point cloud objective quality metrics perform in assessing the quality of learning-based point cloud coding solutions. To achieve this goal, several quality studies were conducted under different viewing conditions and considering several state-of-the-art point cloud coding solutions. Furthermore, extensive objective quality metrics benchmarking was conducted across this doctoral program in order to assess their performance in predicting the quality of learning-based point cloud coding solutions. Ultimately, this led to multiple contributions that were proposed and accepted by the scientific community and that were helpful in understanding the performance of point cloud coding solutions, the impact of the display on quality perception, and the performance of objective point cloud quality metrics.

2025-09-17Doctoral thesis

Open access