Modern Image Quality Assessment

Wang, Zhou; Bovik, A.C.

doi:10.1007/978-3-031-02238-8

Cited by 475 publications

(54 citation statements)

References 130 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…TPDM has been compared with seven state-of-the-art metrics, i.e., the Hausdorff distance (HD) [6,11], the root mean squared error (RMS ) [6,11], 3DW P M 1 and 3DW P M 2 [22], MSDM2 [32], DAME [37] and F M P D [38]. The coherence between the objective values produced by the MVQ metrics and the mean opinion scores (M OS) provided by subjective databases is measured using two different correlation kinds: The Pearson linear correlation coefficient (PLCC or r p ), which measures the prediction accuracy of the objective metrics, and the Spearman rank-order correlation coefficient (SROCC or r s ), which measures the prediction monotonicity [19,34]. Before computing the correlation values, especially the PLCC , it is recommended to conduct a psychometric fitting between the objective scores and the M OS values, in order to partially remove the non-linearity between them.…”

Section: Performance Evaluation and Comparisonsmentioning

confidence: 99%

“…Although during the last decade we have seen tremendous advance in objective image visual quality assessment [19,34], the research on objective mesh visual quality (MVQ) assessment is still at its early stage, with very few metrics proposed [39]. A possible way to evaluate the perceptual quality of 3D meshes is to apply image quality metrics on 2D images generated through 3D model rendering under several pre-selected viewing positions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Curvature-Tensor-Based Perceptual Quality Metric for 3D Triangular Meshes

Torkhani

Wang

Chassery

2014

MG&V

View full text Add to dashboard Cite

Perceptual quality assessment of 3D triangular meshes is crucial for a variety of applications. In this paper, we present a new objective metric for assessing the visual difference between a reference triangular mesh and its distorted version produced by lossy operations, such as noise addition, simplification, compression and watermarking. The proposed metric is based on the measurement of the distance between curvature tensors of the two meshes under comparison. Our algorithm uses not only tensor eigenvalues (i.e., curvature amplitudes) but also tensor eigenvectors (i.e., principal curvature directions) to derive a perceptually-oriented tensor distance. The proposed metric also accounts for the visual masking effect of the human visual system, through a roughness-based weighting of the local tensor distance. A final score that reflects the visual difference between two meshes is obtained via a Minkowski pooling of the weighted local tensor distances over the mesh surface. We validate the performance of our algorithm on four subjectively-rated visual mesh quality databases, and compare the proposed method with state-of-the-art objective metrics. Experimental results show that our approach achieves high correlation between objective scores and subjective assessments.

show abstract

Section: Performance Evaluation and Comparisonsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Curvature-Tensor-Based Perceptual Quality Metric for 3D Triangular Meshes

Torkhani

Wang

Chassery

2014

MG&V

View full text Add to dashboard Cite

show abstract

“…Additionally, objective quality assessment algorithms can be classified according to the availability of an original (distortion-free) image, with which the distorted image is to be compared. Fullreference (FR) methods [8], [12], [14]- [16] directly compare the received and the undistorted (reference) images. In spite of their accuracy [12], [13], FR methods prove impractical when the reference image is unavailable.…”

Section: Introductionmentioning

confidence: 99%

“…In spite of their accuracy [12], [13], FR methods prove impractical when the reference image is unavailable. No-reference (NR) methods [8], [16], [17]- [20] can assess perceived quality without any information about the original image, but are usually targeted to a predefined set of distortions, and therefore their applicability is limited.…”

Section: Introductionmentioning

confidence: 99%

“…In this paper, a reduced reference (RR) approach is preferred. RR methods [8], [16], [21]- [25] are designed to allow quality assessment being supported by only a limited number of numerical features extracted from the original image, paying the feasible additional cost of transmitting side information through the video-chain endpoints [3], [7]. In practice, the resulting small-sized information can be included in the video signal as metadata without major changes in existing broadcasting protocols.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Color Distribution Information for the Reduced-Reference Assessment of Perceived Image Quality

Redi

Gastaldo

Heynderickx

et al. 2010

IEEE Trans. Circuits Syst. Video Technol.

View full text Add to dashboard Cite

Reduced-reference systems can predict in real-time the perceived quality of images for digital broadcasting, only requiring that a limited set of features, extracted from the original undistorted signals, is transmitted together with the image data. The present research uses descriptors based on the color correlogram, analyzing the alterations in the color distribution of an image as a consequence of the occurrence of distortions, for the reduced-reference data. The processing architecture relies on a double layer at the receiver end. The first layer identifies the kind of distortion that may affect the received signal. The second layer deploys a dedicated prediction module for each type of distortion; every predictor yields an objective quality score, thus completing the estimation process. Computational-Intelligence models are used extensively to support both layers with empirical training. The double-layer architecture implements a general-purpose image quality assessment system, not being tied up to specific distortions and, at the same time, it allows to benefit from the accuracy of specific, distortion-targeted metrics. Experimental results based on subjective quality data confirm the general validity of the approach

show abstract