Perceptual Quality Metrics for 3D Meshes: Towards an Optimal Multi-attribute Computational Model

Abstract

Abstract-3D graphical data, commonly represented using triangular meshes, are deployed in a wide range of application processes including compression, filtering, watermarking, and simplification. These processes often introduce geometric distortions which affect the visual quality of the ultimate data visualization. In order to accurately evaluate perceptual impacts caused by the distortions, assessment metrics on 3D Mesh Visual Quality (MVQ) have been extensively discussed in the literature. Researchers recommended various metrics to predict the adverse effects that visual artifacts can have in applications. Most of these metrics are based on geometric attributes, conventional geometric distance, Laplacian coordinates, different types of curvature computation, and dihedral angles. We hypothesize that an optimal combination of multiple attributes associated with a 3D mesh surface can contribute to better perceptual prediction than single attributes used separately. In this paper, we use two user studies to validate our hypothesis. Our contributions are:(1) providing a detailed analysis of the most relevant geometric attributes for mesh quality assessment, and (2) introducing a new perceptual evaluation metric based on multiple attributes, with the optimal combination determined through machine learning techniques. Statistical quantitative analysis shows that our metric delivers better results than other state-of-the-art approaches. The proposed method is simple to implement and fast in execution. Moreover, our framework can easily be expanded to accommodate additional surface attributes.

Metric Evaluation Pipeline for 3d Modeling of Urban Scenes

“…Even small inaccuracies in position can have a noticeable effect on the appearance of a model when rendered with direct lighting or a texture map. Lavoue et al (2013) have assessed computational measures of curvature and roughness similarity and reported that mean curvature and geometric Laplacian metrics correlate well with visual perception metrics. We have implemented the same measures in our pipeline and observed a similar agreement between these two metrics and visual inspection of surface smoothness, as shown in Table 6.…”

“…Source imagery and ground truth data used for experiments have also been publicly released in the Multi-View Stereo 3D Mapping Challenge (MVS3DM) Benchmark (2017). 3D model evaluation metrics in our pipeline include horizontal and vertical accuracy and completeness (similar to metrics employed by Akca et al 2010, Bosch et al 2016, and Sampath et al 2014, volumetric completeness and correctness (similar to work reported by McKeown et al 2000), perceptual quality (based on the work of Lavoue et al 2013), and model simplicity (a relative measure of triangle or geon count). These metrics are intended to expand upon the multiple view stereo analysis by Bosch et al (2016) and enable a comprehensive automated performance evaluation of both geometric and perceptual value of 3D object models reconstructed from imagery as well as assessment of the modeling process at the point cloud reconstruction, semantic labeling, and mesh simplification or model fitting steps.…”

Quality Assessment in Computer Graphics

“…curvature computations [45,42,86], dihedral angles [18,89], Geometric Laplacian [32,84], and Laplacian of Gaussian curvature [97]. Lavoué et al [43] have hypothesized that a combination of these attributes could deliver better results that using them separately. They propose a quality metric based on an optimal linear combination of these attributes determined through machine learning.…”

“…Table 1 summarizes these correlation results; best metrics are highlighted for each database. Note that many metrics cannot be applied to evaluating simplification distortions because they need the compared objects to share the same connectivity - [32,84,45,89,43] -or the same level of details - [18]. Table 1 Correlation between Mean Opinion Scores and values from the metrics for four publiclyavailable subjective databases.…”

A perceptual quality metric for 3D triangle meshes based on spatial pooling

“…The image quality assessment community has conducted significant studies on using machine learning techniques to support IVQ assessment [11]- [17], while most existing mesh quality assessment approaches [1]- [6] attempt to model the HVS explicitly. Lavoue et al [18] proposed a perceptual evaluation metric based on multiple attributes using machine learning techniques. This metric did not take the characteristics of HVS into account, and cross-database performance was not presented to justify its generalization capability.…”