Gradient Magnitude Similarity Deviation: A Highly Efficient Perceptual Image Quality Index

By Wufeng Xue, Lei Zhang, Xuanqin Mou, A. Bovik. IEEE Transactions on Image Processing, 2013

1,637 citations recorded by the source corpus.

It is an important task to faithfully evaluate the perceptual quality of output images in many applications, such as image compression, image restoration, and multimedia streaming. A good image quality assessment (IQA) model should not only deliver high quality prediction accuracy, but also be computationally efficient. The efficiency of IQA metrics is becoming particularly important due to the increasing proliferation of high-volume visual data in high-speed networks. We present a new effective and efficient IQA model, called gradient magnitude similarity deviation (GMSD). The image gradients are sensitive to image distortions, while different local structures in a distorted image suffer different degrees of degradations. This motivates us to explore the use of global variation of gradient based local quality map for overall image quality prediction. We find that the pixel-wise gradient magnitude similarity (GMS) between the reference and distorted images combined with a novel pooling strategy-the standard deviation of the GMS map-can predict accurately perceptual image quality. The resulting GMSD algorithm is much faster than most state-of-the-art IQA methods, and delivers highly competitive prediction accuracy. MATLAB source code of GMSD can be downloaded at http://www4.comp.polyu.edu.hk/~cslzhang/IQA/GMSD/GMSD.htm.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/91a73e9c6cbba422ee50e287f1bfa9ba30f6922b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

A Patch-Structure Representation Method for Quality Assessment of Contrast Changed Images (2015, methodology): “The PCQI map produced by the proposed method also indicates local quality variations, which are useful in guiding the future design of advanced image enhancement algorithms.”
TOPIQ: A Top-Down Approach From Semantics to Distortions for Image Quality Assessment (2023, positive): “To address this, Wang et al. [13] introduced the structural similarity (SSIM) index to compare structural similarity in local patches, which inspired a lot of follow-up works [7], [14], [15], [16], [17], [18], [19].”
Blind Image Quality Assessment Based on High Order Statistics Aggregation (2016, methodology): “Typical FR and RR methods include SSIM [3], FSIM [4], RRED [5] and GMSD [6].”
Using labeled data to evaluate change detectors in a multivariate streaming environment (2009, positive): “EDANet [17] also utilizes the strategy of dense connection, and the difference is that EDANet adopts the dense connection between modules using asymmetric convolution.”
dipIQ: Blind Image Quality Assessment by Learning-to-Rank Discriminable Image Pairs (2017, methodology): “We ﬁrst choose three best-trusted FR-IQA models, namely MS-SSIM [82], VIF [83], and GSMD [84].”
A fusion-based enhancing method for weakly illuminated images (2016, methodology): “For the full reference IQA we use the gradient magnitude similarity deviation (GMSD) metric [41] to evaluate visual distortion between the original and enhanced images.”
KADID-10k: A Large-scale Artificially Distorted IQA Database (2019, neutral): “665 GMSD [19] 0.”
UGC-VQA: Benchmarking Blind Video Quality Assessment for User Generated Content (2020, methodology): “Inspired by the efﬁcacy of standard deviation pooling as ﬁrst introduced in GMSD [71] and later also used in TLVQM [40], we calculate these spatial features every second frame within each sequentially cut non-overlapping 1-second chunk, then we enrich the feature set by applying average and standard”
Deep Neural Networks for No-Reference and Full-Reference Image Quality Assessment (2016, methodology): “Dropout regularization with a ratio of 0.5 is applied to the fully-connected layers in order to prevent overﬁtting in the regression [24].”
Loss Functions for Image Restoration With Neural Networks (2017, methodology): “The Gradient Magnitude Similarity Deviation (GMSD) is characterized by simplified math and performance similar to that of SSIM, but it requires computing the standard deviation over the whole image [22].”
From Patches to Pictures (PaQ-2-PiQ): Mapping the Perceptual Space of Picture Quality (2019, background): “Full-Reference models: Many full-reference (FR) perceptual picture quality predictors, which make comparisons against high-quality reference pictures, are available [5, 6], [27, 28, 29, 30, 31, 32, 33].”
PieAPP: Perceptual Image-Error Assessment Through Pairwise Preference (2018, methodology): “To compare, we also test 1) model-based methods: Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), SSIM [52], MS-SSIM [53], GMSD [54], VSI [56], PSNR-HMA [41], FSIMc [57], SFF [12], and SCQI [4], and 2) learning-based methods: DOG-SSIMc [40], Lukin et al. [35], Kim et al. [25], and Bosse et”
Loss Functions for Neural Networks for Image Processing (2015, methodology): “The Gradient Magnitude Similarity Deviation (GMSD) is characterized by simpliﬁed math and performance similar to that of SSIM, but it requires computing the standard deviation over the whole image, and therefore cannot be computed on a single patch [22].”
Image Quality Assessment: Unifying Structure and Texture Similarity (2020, methodology): “The SSIM index [2] introduced the concept of comparing structure similarity (instead of measuring error visibility), opening the door to a new class of full-reference IQA measures [16], [23]–[25].”
An Experiment-Based Review of Low-Light Image Enhancement Methods (2020, methodology): “…the patch-based contrast quality index (PCQI) [302], the colorfulness-based PCQI (CPCQI) [303], the Gradient Magnitude Similarity Deviation (GMSD) [304], the visual information ﬁdelity (VIF) [305], the visual saliency index (VSI) [306], the tone-mapped image quality index (TMQI) [307], the…”
Blind Image Quality Estimation via Distortion Aggravation (2018, methodology): “GSI [24], FSIM [25], GMSD [26], PSIM [14].”