Davi Lazzarotto’s research while affiliated with École Polytechnique Fédérale de Lausanne and other places

... Numerous studies have been conducted to evaluate the quality of point clouds, taking into account several coding approaches and experimental configurations [9,10,11,4,12,13]. Perry et al. presented an assessment of the perceived quality of MPEG Point Cloud codecs, notably Video Point Cloud Compression (V-PCC) and Geometry Point Cloud Compression (G-PCC), using a 2D display [2]. ...

Reference:

Quality Analysis of the Coding Bitrate Tradeoff Between Geometry and Attributes for Colored Point Clouds

... Due to this stability, some FR-PCQA metrics [17,18] have been employed as criteria for establishing effective compression methods in the MPEG standardization [5]. While these methods [17,18] focus on pointwise errors, other methods that consider (i) more complex features (e.g., structural similarity [19,20,21]), (ii) graph similarity [22,23], or (iii) learning-based techniques [24,25] have been proposed to improve the PCQA accuracy. ...

Reference:

Full reference point cloud quality assessment using support vector regression

... This will include an adaptation to encode LiDAR PCs in both Cartesian and cylindrical coordinates. Additionally, among other advancements, recent developments have introduced sparse convolutions to PC compression [40]. ...

Reference:

Impact of LiDAR point cloud compression on 3D object detection evaluated on the KITTI dataset

... Traditional subjective quality assessment techniques, like those presented in ITU-T Recommendation BT.500 [1] and reviewed in Part 1 of the JPEG AIC standard [2], are often effective for evaluating images with low and medium visual quality. However, when compared to quality scale reconstruction from pair comparisons, they lack precision [3], and they fall short when adopted to evaluate the visual quality of high-fidelity contents, which requires distinguishing images with subtle variations in visual quality [4]. For these reasons, the JPEG Committee launched a new activity in 2021, known as JPEG AIC-3 [5], with the goal of a fine-grained quality assessment of compressed images with high-fidelity. ...

Reference:

Fine-grained subjective visual quality assessment for high-fidelity compressed images

... Five images from the JPEG AIC-3 dataset were selected to represent a diverse range of image types and content and cropped to a size of 620 × 800 pixels, as presented in Figure 2. These images had been compressed with five codes, JPEG, JPEG 2000, VVC Intra, JPEG XL, and AVIF at 10 bitrates each, corresponding approximately to JND values equally spaced from 0.25 to 2.5, as determined by a pairwise comparison experiment [14]. ...

Reference:

Fine-grained subjective visual quality assessment for high-fidelity compressed images

... Avatar-based communication has been considered in [16], where the point cloud of avatars, structures, and models are transmitted between transmitter and receiver. Task-related effectiveness level performance metrics, including point-to-point [17], peak signal-to-noise ratio for the luminance component [18], mean per joint position error [19] have been considered to assess the telepresence task [20], point cloud video displaying task [21], and avatar pose recovery task [22], respectively. However, these AR-related applications have not fully addressed the issue of the effectiveness of avatar transmission, and bandwidth requirements for such applications still remain high. ...

Reference:

Task-oriented and Semantics-aware Communication Framework for Augmented Reality

... Moreover, a subjective quality evaluation targeting machine-learning-based coding solutions was reported [17]. In early 2022, a quality assessment study was performed to support the JPEG Pleno point cloud coding CfP [28]. This study aimed to evaluate the current state-of-the-art point cloud solutions, analyze the stability of the subjective quality assessment methodologies, and evaluate the performance of objective metrics. ...

Reference:

Quality evaluation of point cloud compression techniques

... A first study conducted a crowdsourced evaluation [37] of G-PCC, V-PCC, and two learning-based methods [15,16]. The same author also used a learning-based coding tool [18] and G-PCC in an evaluation [38] with both a flat screen and a light field monitor. Recently, a large-scale study [39] produced subjective scores for a set of more than 1200 distorted point clouds using G-PCC, V-PCC, and a learning-based algorithm for geometry compression [15], with the goal of fostering research on learning-based objective quality metrics. ...

Reference:

Subjective performance evaluation of bitrate allocation strategies for MPEG and JPEG Pleno point cloud compression

... (A) Average results over the kodak dataset. Our novel SDC coding scheme overperforms all the state of the art coders by at least 0.5 to 3 dB (JPEG DNA BC:[2], JPEG DNA BC Transcoder:[18], JPEG DNA SFC4:[17]), (B) Average result curve over kodak dataset, the MDC side curve is the mean curve across different descriptions. The benchmark is done with the following configuration: N number of descriptions with N = {2, 4}, α = 0.1. ...

Reference:

Implicit Neural Multiple Description for DNA-Based Data Storage

... This residual coding neural network consists of an autoencoder, quantization, and a variational hyperprior for adaptive entropy coding architecture, similar to [29], but the decoder additionally includes a CNN-based refiner module that receives the output of the autoencoder and the distorted PC from the base layer, producing the enhancement layer PC. Frank et al. [34] later proposed a latent space slicing approach to improve the entropy coding in the commonly adopted autoencoder with variational hyperprior architecture. In the proposed approach, the latent representation produced by the autoencoder is sliced along the channel dimension, with each slice being separately entropy coded. ...

Reference:

The JPEG Pleno Learning-based Point Cloud Coding Standard: Serving Man and Machine