A convolutional neural network framework for blind mesh visual quality assessment

Abouelaziz, Ilyass; Cherifi, Hocine

doi:10.1109/icip.2017.8296382

Cited by 23 publications

(13 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• 3D VQA metrics: No-reference quality assessment metrics designed especially for 3D meshes usually without color: NR-SVR [8], NR-GRNN [9], NR-CNN [10].…”

Section: Evaluation Competitorsmentioning

confidence: 99%

“…Currently speaking, most model-based metrics are full-reference (FR) models and they are inspired by similar ideas exploited in IQA methods like mean squared error (MSE) at the vertex level, distribution distance computed between feature maps and structure similarity based on geometry. Considering that the reference is not always available, some NR model-based metrics are proposed, which are mainly developed on the basis of geometry characteristics such as curvature, dihedral angle, and roughness [8] [9] [10]. Such metrics are able to predict the quality of 3D meshes only when the geometry distortions exist, and they might fail to evaluate the visual quality of 3D meshes with color distortions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A No-Reference Visual Quality Metric For 3D Color Meshes

Zhang

Sun

Min

et al. 2021

2021 IEEE International Conference on Multimedia &Amp; Expo Workshops (ICMEW)

View full text Add to dashboard Cite

To improve the viewers' quality of experience (QoE) in computer graphics applications, the visual quality assessment (VQA) of 3D meshes is becoming a popular task in the multimedia area. Since 3D meshes are quite sensitive to the processing operations like simplification and compression, many studies concerning the VQA of 3D meshes have been carried out to measure the caused degradations. However, the previous studies mostly utilize full-reference metrics and focus mainly on the geometry attributes. While in some application scenarios such as 3D reconstruction, digital entertainment, and medical modeling, the reference 3D mesh is not always available and the color information can not be ignored. Therefore, in this paper, we propose a no-reference visual quality metric for 3D color meshes, which is based on the statistical parameters and entropy estimated from the probability distributions of curvature, dihedral angles, face area, face angle, and diffuse color information. The performance of the proposed method is validated on a database specially built for VQA of 3D color meshes. Experimental results show that our metric operates stably and achieves the highest correlation with subjective judgement.

show abstract

“…• 3D VQA metrics: No-reference quality assessment metrics designed especially for 3D meshes usually without color: NR-SVR [8], NR-GRNN [9], NR-CNN [10].…”

Section: Evaluation Competitorsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A No-Reference Visual Quality Metric For 3D Color Meshes

Zhang

Sun

Min

et al. 2021

2021 IEEE International Conference on Multimedia &Amp; Expo Workshops (ICMEW)

View full text Add to dashboard Cite

show abstract

“…2) The Development of MQA: The MQA metrics can be categorized into two types: model-based metrics [11] [12] [16] [23] [24] which operate directly on the 3D models, and image quality assessment (IQA) based metrics [18] [19] [25] [26] [27] which operate on the rendering snapshots of 3D models.…”

Section: A Previous 3d-qa Workmentioning

confidence: 99%

No-Reference Quality Assessment for 3D Colored Point Cloud and Mesh Models

Zhang,

Sun,

Min

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…It also uses SVR for feature vector extraction with a specific end goal to get last target quality score. A deep learning approach was proposed by [34] employing multilayer CNN with multilayer max-pooling. Similar to previous, it also considers curvature and dihedral angles from each distorted mesh.…”

Section: E Blind Evaluationmentioning

confidence: 99%

Perceptual Quality Evaluation of 3D Triangle Mesh: A Technical Review

Muzahid

Wan

Feng

2018

2018 International Conference on Audio, Language and Image Processing (ICALIP)

View full text Add to dashboard Cite

During mesh processing operations (e.g. simplifications, compression, and watermarking), a 3D triangle mesh is subject to various visible distortions on mesh surface which result in a need to estimate visual quality. The necessity of perceptual quality evaluation is already established, as in most cases, human beings are the end users of 3D meshes. To measure such kinds of distortions, the metrics that consider geometric measures integrating human visual system (HVS) is called perceptual quality metrics. In this paper, we direct an expansive study on 3D mesh quality evaluation mostly focusing on recently proposed perceptual based metrics. We limit our study on greyscale static mesh evaluation and attempt to figure out the most workable method for real-time evaluation by making a quantitative comparison. This paper also discusses in detail how to evaluate objective metric's performance with existing subjective databases. In this work, we likewise research the utilization of the psychometric function to expel non-linearity between subjective and objective values. Finally, we draw a comparison among some selected quality metrics and it shows that curvature tensor based quality metrics predicts consistent result in terms of correlation.

show abstract

A convolutional neural network framework for blind mesh visual quality assessment

Cited by 23 publications

References 15 publications

A No-Reference Visual Quality Metric For 3D Color Meshes

A No-Reference Visual Quality Metric For 3D Color Meshes

No-Reference Quality Assessment for 3D Colored Point Cloud and Mesh Models

Perceptual Quality Evaluation of 3D Triangle Mesh: A Technical Review

Contact Info

Product

Resources

About