Multiscale structural similarity for image quality assessment
The DOI matches its Crossref record; 16 of the 16 references it cites were found in a registry or an index; none of the 14 sources its claims cite is checked for support yet.
Abstract
The structural similarity image quality paradigm is based on the assumption that the human visual system is highly adapted for extracting structural information from the scene, and therefore a measure of structural similarity can provide a good approximation to perceived image quality. This paper proposes a multiscale structural similarity method, which supplies more flexibility than previous single-scale methods in incorporating the variations of viewing conditions. We develop an image synthesis method to calibrate the parameters that define the relative importance of different scales. Experimental comparisons demonstrate the effectiveness of the proposed method.
Overview
CoChat Summary
TL;DR
This paper proposes a multi-scale structural similarity (MS-SSIM) image quality method that combines contrast and structure comparisons across scales and luminance comparison at the coarsest scale. On the LIVE JPEG/JPEG2000 database, MS-SSIM outperformed PSNR, the Sarnoff model, and single-scale SSIM variants on all reported evaluation criteria.
Key findings
- MS-SSIM achieved the best reported performance: CC 0.969, ROCC 0.966, MAE 3.86, RMS 4.91, and OR 1.16%, compared with the best single-scale SSIM at M=2 (CC 0.963, ROCC 0.959, MAE 4.21, RMS 5.38, OR 2.62%) and PSNR (CC 0.905, ROCC 0.901, MAE 6.53, RMS 8.45, OR 15.7%).
- Single-scale SSIM performance varied with scale; the best single-scale result was at M=2, and single-scale SSIM tended to give higher scores as scale increased.
- The calibrated cross-scale exponents were β1=γ1=0.0448, β2=γ2=0.2856, β3=γ3=0.3001, β4=γ4=0.2363, and α5=β5=γ5=0.1333.
- The authors conclude that multi-scale SSIM provides a meaningful balance between scales and outperforms the best single-scale SSIM and state-of-the-art image quality metrics.
Methods
- The MS-SSIM system iteratively applies a low-pass filter and downsamples by 2; at each scale it computes contrast and structure comparisons, computes luminance comparison only at Scale M, and combines them using exponents normalized so that ∑γj=1.
- Cross-scale parameters were calibrated by image synthesis: 5 scales and 12 distortion levels (range 2^3 to 2^14) were used with 10 original 64×64 images, producing 600 distorted images; 8 subjects selected images of equal perceived quality across scales at a viewing distance of 32 pixels per degree of visual angle.
- Evaluation used the LIVE database with 344 JPEG and JPEG2000 compressed images (bit rate 0.028 to 3.150 bits/pixel), MOS from 13–25 subjective scores, and compared PSNR, Sarnoff JNDmetrix 8.0, single-scale SSIM for M=1–5, and MS-SSIM using CC, MAE, RMS, OR, and ROCC after logistic mapping.
Limitations
- Stated: the image-synthesis calibration approach is “still rather crude,” and the authors are working on a more systematic approach.
- Stated: CSF cannot be directly used to derive the parameters because it is typically measured at visibility threshold with simplified stimuli, whereas the goal is comparing complex structured images at visible distortion levels.
- Evident from the text: calibration used only 10 original 64×64 grayscale images and 8 subjects, including one author, and evaluation was limited to the LIVE JPEG/JPEG2000 database.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 1 Not valid · 14 Not checked · 18 Own results and interpretations
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“It can also be observed that the single-scale model tends to supply higher scores with the increase of scales.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Finally, for every one of the objective evaluation criteria, multi-scale SSIM model outperforms all the other models, including the best single-scale SSIM model, suggesting a meaningful balance between scales.”
- Also stated in: § 5. DISCUSSIONS · p. 4 · hedged
“Experiments show that with an appropriate parameter settings, the multi-scale method outperforms the best single-scale SSIM model as well as state-of-the-art image quality metrics.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 5. DISCUSSIONS · p. 4 · hedged
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In general, the subjects agreed with each other on each image more than they agreed with themselves across different images.”
plainstates an associationn = 8- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The resulting parameters we obtained are β 1 = γ 1 = 0.0448, β 2 = γ 2 = 0.2856, β 3 = γ 3 = 0.3001, β 4 = γ 4 = 0.2363, and α 5 = β 5 = γ 5 = 0.1333, respectively.”
plainn = 8- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“However, the SSIM index algorithm introduced in [5] is a single-scale approach.”
What the sources say
- Not matchedcites [5] [5] Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, 'Image quality assessment: From error measurement to structural similarity,' IEEE Trans. Image Processing , vol. 13, Jan. 2004. -
More than one paper fits this citation.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
10 more claims
Create a free account to see all 33 claims and every source verdict.
- It may state two separate claims as one. 8
- Probably not a substantial claim. 13
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Further, as research in quality assessment has demonstrated, incorporating multiscale information when assessing quality produces QA algorithms that perform better in terms of correlation with human perception [13], [29].”
From No-Reference Image Quality Assessment in the Spatial Domain (2012) · cites this paper as [13]
Citations
Cites
Image quality assessment: from error visibility to structural similarity
Wang, Z., Bovik, A. C., Sheikh, H. R., & Simoncelli, E. P. (2004) · IEEE Transactions on Image Processing · doi:10.1109/tip.2003.819861 · Cited by 57,892
Self-citationA universal image quality index
Wang, Z., & Bovik, A. C. (2002) · IEEE Signal Processing Letters · doi:10.1109/97.995823 · Cited by 5,757
Self-citationImage quality measures and their performance
Eskicioglu, A. M., & Fisher, P. S. (1995) · IEEE Transactions on Communications · doi:10.1109/26.477498 · Cited by 2,117 · cited 1×
“Traditional perceptual image quality assessment methods are based on a bottom-up approach which attempts to simulate the functionality of the relevant early human visual system (HVS) components.”
BackgroundA Parametric Texture Model Based on Joint Statistics of Complex Wavelet Coefficients
Portilla, J., & Simoncelli, E. P. (2000) · International Journal of Computer Vision · doi:10.1023/a:1026553619983 · Cited by 1,806 · cited 2×
“These synthesis methods provide a powerful and efficient means of testing a model, and have the added benefit that the resulting images suggest improvements that might be made to the model [11].”
MethodSelf-citation(1995) · Medical Entomology and Zoology · Cited by 1,212
Pyramid-based texture analysis/synthesis
Heeger, D. J., & Bergen, J. R. (1995) · Proceedings of the 22nd annual conference on Computer graphics and interactive techniques - SIGGRAPH '95 · doi:10.1145/218380.218446 · Cited by 1,095
Video quality assessment based on structural distortion measurement
Wang, Z., Lu, L., & Bovik, A. C. (2004) · Signal Processing: Image Communication · doi:10.1016/s0923-5965(03)00076-6 · Cited by 1,024
Self-citationImage Quality Assessment: From Error Measurement to Structural Similarity
Wang, Z., Bovik, A. C., Sheikh, H. R., & Simoncelli, E. P. (2004) · Cited by 919
Self-citationTeo, P. C., & Heeger, D. J. (2002) · Proceedings of 1st International Conference on Image Processing · doi:10.1109/icip.1994.413502 · Cited by 471
Objective video quality assessment
(2003) · Handbook of Video Databases · doi:10.1201/9780203489864-45 · Cited by 359
Research works printed but not linked: 2
- 0, JNDmetrix Technology, Sarnoff Corporation
- 2000, Final report from the video quality experts group on the validation of objective models of video quality assessment, VQEG
Cited by
Items 2,101–2,200 of 6,047
Cuenca, E. C., R., M. A. S., & Guerrero-Bueno, Z. (2020) · Journal of Archaeological Science · doi:10.1016/j.jas.2020.105269 · Cited by 17
Deep Arbitrary HDRI: Inverse Tone Mapping With Controllable Exposure Changes
Jo, S. Y., Lee, S., Ahn, N., & Kang, S. (2021) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2021.3087034 · Cited by 14
Automatic consecutive context perceived transformer GAN for serial sectioning image blind inpainting
Wang, L., Zhang, S., Gu, L., Zhang, J., Zhai, X., Sha, X., & Chang, S. (2021) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2021.104751 · Cited by 15
Towards Adversarially Robust Deep Image Denoising
Yan, H., Zhang, J., Feng, J., Sugiyama, M., & Tan, V. Y. F. (2022) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2022/211 · Cited by 23
Luo, S. (2021) · International Conference on Innovation in Artificial Intelligence (ICIAI) · doi:10.1145/3461353.3461388 · Cited by 16
“MS-SSIM [80], multi-scale SSIM, a more advanced form of SSIM, is conducted over multiple scales through a process of multiple stages of sub-sampling.”
MethodInfluentialJin, D., Zheng, H., Zhao, Q., Wang, C., Zhang, M., & Yuan, H. (2021) · Tomography · doi:10.3390/tomography7040064 · Cited by 12
Pix2Pix Network to Estimate Agricultural Near Infrared Images from RGB Data
de Lima, D. C., Saqui, D., Mpinda, S. A. T., & Saito, J. H. (2022) · Canadian Journal of Remote Sensing · doi:10.1080/07038992.2021.2016056 · Cited by 14
“This network was developed to deal with unpaired images, in other words, images that do not have matching pixels with their respective ground truths, so it is an unsupervised image translation model, such as the image-to-image translation models described in Wang et al. (2020). The cycleGAN network is widely used in s…”
BackgroundInfluentialQRnet: fast learning-based QR code image embedding
Pena-Pena, K., Lau, D. L., Arce, A. J., et al. (2022) · Multimedia Tools and Applications · doi:10.1007/s11042-022-12357-6 · Cited by 13
“In Table 3, the PSNR, MSSIM, J, SSIM [27], and NCC [5] were used to measure the visual quality of the QR Images generated by each method with respect to the original color image.”
MethodA Multi-Stage Visible and Infrared Image Fusion Network Based on Attention Mechanism
Zheng, X., Yang, Q., Si, P., & Wu, Q. (2022) · Sensors · doi:10.3390/s22103651 · Cited by 13
“To prevent the training bias caused by manual weighted loss functions, we designed a novel Pull–Push loss function based on the weighted edge information evaluation index [40], combined with the widely used SSIM loss function [41].”
MethodOoDHDR-Codec: Out-of-Distribution Generalization for HDR Image Compression
Cao, L., Jiang, A., Li, W., Wu, H., & Ye, N. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i1.19890 · Cited by 17
“For distortion measurement, MS-SSIM metrics (Wang, Simoncelli, and Bovik 2003) is employed in e sdr and the pixel-wise L 1 loss (MAE) is employed in e hdr .”
MethodAugmented Reality Image Quality Assessment Based on Visual Confusion Theory
Duan, H., Guo, L., Sun, W., Min, X., Chen, L., & Zhai, G. (2022) · IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB) · doi:10.1109/bmsb55706.2022.9828671 · Cited by 20
Коваленко, Б., Лукин, В. В., Kryvenko, S., et al. (2022) · Applied Sciences · doi:10.3390/app12157555 · Cited by 13
“The plots presented in Figure 6 for noise”
BackgroundVarga, D. (2022) · Journal of Imaging · doi:10.3390/jimaging8080224 · Cited by 13
“Namely, the author proposed a combined metric using the product and power of MS-SSIM [34], VIF [38], and R-SVD [47].”
MethodInfluentialSSIM-Variation-Based Complexity Optimization for Versatile Video Coding
Lin, J., Lin, H., Zhang, Z., et al. (2022) · IEEE Signal Processing Letters · doi:10.1109/lsp.2022.3227748 · Cited by 13
“The above results indicate that the proposed method with SSIM outperforms the variation of PSNR and MS-SSIM.”
MethodAdaptive-Attention Completing Network for Remote Sensing Image
Huang, W., Deng, Y., Hui, S., & Wang, J. (2023) · Remote Sensing · doi:10.3390/rs15051321 · Cited by 14
Коваленко, Б., Лукин, В. В., & Vozel, B. (2023) · Remote Sensing · doi:10.3390/rs15061669 · Cited by 13
PAR2Net: End-to-End Panoramic Image Reflection Removal
Hong, Y., Zheng, Q., Zhao, L., Jiang, X., Kot, A. C., & Shi, B. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3286429 · Cited by 13
“TABLE I COMPARISON OF QUANTITATIVE RESULTS IN TERMS OF PSNR [14] AND SSIM [53] ON THE PORTABLE DATASET FOR EVALUATING THE RECOVERY OF BOTH TRANSMISSION SCENES AND REFLECTION LAYERS”
BackgroundMethodInfluentialADD-UNet: An Adjacent Dual-Decoder UNet for SAR-to-Optical Translation
Luo, Q., Li, H., Chen, Z., & Li, J. (2023) · Remote Sensing · doi:10.3390/rs15123125 · Cited by 14
SCAUIE-Net: Underwater Image Enhancement Method Based on Spatial and Channel Attention
Zhong, Y., Wang, J., & Lu, Q. (2023) · IEEE Access · doi:10.1109/access.2023.3291449 · Cited by 13
“Due to the human visual system being more sensitive to luminance and color variations in texture-less regions, SCAUIE-Net uses a loss function mixed with Perceptual Loss [7], MS-SSIM Loss [8] and MAE Loss.”
MethodIDGI: A Framework to Eliminate Explanation Noise from Integrated Gradients
Yang, R., Wang, B., & Bilgic, M. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02272 · Cited by 38
“We also introduce a modified version of SIC and AIC with MS-SSIM information level.”
MethodInfluentialStructure-Aware Lighting Design for Volume Visualization
Tao, Y., Lin, H. B., Dong, F., et al. (2012) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2012.267 · Cited by 12
Objective evaluation of naturalness, contrast, and colorfulness of tone-mapped images
Krasula, L., Fliegel, K., Callet, P. L., & Klíma, M. (2014) · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · doi:10.1117/12.2075270 · Cited by 16
Denoising and the active warden
Smith, C. B., & Agaian, S. С. (2007) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/icsmc.2007.4413746 · Cited by 15
“The SSIM is a measure based on the idea that the human visual system appears to capture and use structural information more than other image features [7].”
BackgroundMultisensor Super Resolution Using Directionally-Adaptive Regularization for UAV Images
Kang, W., Yu, S., Ko, S., & Paik, J. (2015) · Sensors · doi:10.3390/s150512053 · Cited by 14
LR-GAN for degraded Face Recognition
Banerjee, S., & Das, S. (2018) · Pattern Recognition Letters · doi:10.1016/j.patrec.2018.10.034 · Cited by 15
A Perceptual Distinguishability Predictor For JND-Noise-Contaminated Images
Hadizadeh, H., Heravi, A. R., Bajić, I. V., & Karami, P. (2018) · IEEE Transactions on Image Processing · doi:10.1109/tip.2018.2883893 · Cited by 13
“For this purpose, we used the popular MS-SSIM metric [28].”
MethodMehrish, A., Subramanyam, A., & Emmanuel, S. (2019) · IEEE Access · doi:10.1109/access.2019.2901345 · Cited by 13
“A similar observation can be made for other quality measures (MS-SSIM, PSNRHSV, VIF).”
MethodContinuous Prediction for Quality of Experience in Wireless Video Streaming
Shi, W.-J., Sun, Y., & Pan, J. (2019) · IEEE Access · doi:10.1109/access.2019.2919610 · Cited by 13
“In this paper, FR image quality assessment (IQA) methods PSNR, SSIM [29] and MS-SSIM [30], RR IQA methods STRRED [13] and OSVPH, NR IQA method NIQE [31] are used as assessment method for FQ.”
MethodUsman, M. A., & Martini, M. G. (2019) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2019.103383 · Cited by 13
Hiding Video in Audio via Reversible Generative Models
Yang, H., Ouyang, H., Koltun, V., & Chen, Q. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00119 · Cited by 16
“We evaluate the change in image quality using MS-SSIM [38].”
MethodBVI-SynTex: A Synthetic Video Texture Dataset for Video Compression and Quality Assessment
Katsenou, A., Dimitrov, G., Ma, D., & Bull, D. (2020) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2020.2976591 · Cited by 14
End-to-End Learning for Video Frame Compression with Self-Attention
Zou, N., Zhang, H., Cricri, F., Tavakoli, H. R., Lainema, J., Aksu, E., Hannuksela, M. M., & Rahtu, E. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00079 · Cited by 15
“We train our model by using the following objective: (11) where D 1 is the negative multi-scale structural similarity (MS-SSIM) [14], D 2 is the mean-squared error (MSE), R is the rate-loss provided by the probability model, M is the constraint on the importance map defined in Eq.”
InfluentialVisual Quality Assessment for Interpolated Slow-Motion Videos Based on a Novel Database
Men, H., Hosu, V., Lin, H., Bruhn, A., & Saupe, D. (2020) · International Conference on Quality of Multimedia Experience (QoMEX) · doi:10.1109/qomex48832.2020.9123096 · Cited by 20
“476 MS-SSIM [51] 0.”
Learning to Improve Image Compression Without Changing the Standard Decoder
Strümpler, Y., Yang, R., & Timofte, R. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-66823-5_12 · Cited by 18
“To better represent local statistics we also use the Multi-Scale Structural Similarity (MS-SSIM) [35] with its default implementation tf.”
MethodLearning an adaptive model for extreme low‐light raw image processing
Fu, Q., Di, X., & Zhang, Y. (2020) · IET Image Processing · doi:10.1049/iet-ipr.2020.0100 · Cited by 17
“By introducing SSIM into loss function, our method shows more promising results in images with text and has advantages over SID in keeping more edge information, e.g. the images in the first row of Fig.”
BackgroundMethodInfluentialA Gated and Bifurcated Stacked U-Net Module for Document Image Dewarping
Bandyopadhyay, H., Dasgupta, T., Das, N., & Nasipuri, M. (2021) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr48806.2021.9413001 · Cited by 16
“1333 respectively following the original implementation [30].”
Image Quality Assessment using Synthetic Images
Madhusudana, P. C., Birkbeck, N., Wang, Y., Adsumilli, B., & Bovik, A. C. (2022) · IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw54805.2022.00015 · Cited by 17
Self-citationStudy of subjective and objective quality assessment of infrared compressed images
Zelmati, O., Bondžulić, B., Pavlović, B. Z., Tot, I., & Merrouche, S. (2022) · Journal of Electrical Engineering · doi:10.2478/jee-2022-0011 · Cited by 13
Maliamanis, T. V., Apostolidis, K. D., & Papakostas, G. A. (2022) · Biomedicines · doi:10.3390/biomedicines10102545 · Cited by 14
“While SSIM drops, PGD, FGSM and Square Attack perform better than Mb-AdA.”
MethodInfluentialA Subjective Quality Study for Video Frame Interpolation
Danier, D., Zhang, F., & Bull, D. (2022) · Proceedings - International Conference on Image Processing · doi:10.1109/icip46576.2022.9897364 · Cited by 18
“The DMOS and metric values and the fitted logistic curves for PSNR, SSIM, LPIPS and VMAF are also plotted in Fig.”
MethodInfluentialA Multi-scale Structure SIMilarity metric for image fusion qulity assessment
Xiao, Z.-S. (2011) · International Conference on Wavelet Analysis and Pattern Recognition · doi:10.1109/icwapr.2011.6014491 · Cited by 15
Quality-Driven Joint Rate and Power Adaptation for Scalable Video Transmissions Over MIMO Systems
Chen, X., Hwang, J., Ritcey, J. A., et al. (2015) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2015.2511815 · Cited by 13
“Among these VQA schemes, the multiscale structural similarity index (MS-SSIM) developed in [29], originally designed as an image quality assessment method, has been reported as a simple but effective VQA scheme, which has high correlation with subjectively perceived video quality [30]–[32].”
MethodMeasurement of critical temporal inconsistency for quality assessment of synthesized video
Kim, H. G., & Ro, Y. M. (2016) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2016.7532513 · Cited by 12
Content-Sensitive Data Compression for IoT Streaming Services
Hsu, C.-C., Fang, Y.-T., & Yu, F. (2017) · IEEE International Congress on Internet of Things (ICIOT) · doi:10.1109/ieee.iciot.2017.25 · Cited by 13
“Unlike [10] where the weight of α (luminance), β (contrast), and γ (structure) is set 1:1:1, we adjust the weights to various ratios to explore how different ratios affect the compression.”
BackgroundMethodInfluentialcencro - Speedup of Video Quality Calculation using Center Cropping
Göring, S., Krammer, C., & Raake, A. (2019) · IEEE International Symposium on Multimedia (ISM) · doi:10.1109/ism46123.2019.00010 · Cited by 13
“On the other side, classical FR models are based on measuring differences between a distorted video/image and a reference video/image, e.g. PSNR, MS-SSIM [37, 38] or VIF [33].”
BackgroundDeng, M., Li, S., & Barbastathis, G. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.07945 · Cited by 15
“This again verifies the well-known fact that neither PSNR nor SSIM is a good performance metric for visual quality[19, 41, 42, 21, 24].”
BackgroundAnonymousNet: Natural Face De-Identification with Measurable PrivacyPreprint
Li, T., & Lin, L. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1904.12620 · Cited by 13
“7894 MS-SSIM [59] 0.”
Teacher-Student Compression with Generative Adversarial NetworksPreprint
Liu, R., Fusi, N., & Mackey, L. W. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1812.02271 · Cited by 27
“Some standard quality measures, like multiscale structural similarity (Wang et al., 2003), quantify the diversity of a synthetic dataset but do not capture class affinity, the ability of datapoints to be correctly associated with their labels with high confidence.”
BackgroundLee, J., Cho, S.-H., & Kim, M. (2019) · arXiv (Cornell University) · Cited by 12
Dense Dilated UNet: Deep Learning for 3D Photoacoustic Tomography Image ReconstructionPreprint
Guan, S., Hsu, K.-T., Eyassu, M., & Chitnis, P. V. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2104.03130 · Cited by 13
“To evaluate image reconstruction quality, the multi-scale structural similarity index metric (MS-SSIM) was used to compare the reconstructed image to the ground truth image [47].”
Zhang, Z., Zhou, Y., & Ma, J. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.126251 · Cited by 15
SwiftSketch: A Diffusion Model for Image-to-Vector Sketch Generation
Arar, E., Frenkel, Y., Cohen‐Or, D., Shamir, A., & Vinker, Y. (2025) · ACM SIGGRAPH Conference (SIGGRAPH) · doi:10.1145/3721238.3730612 · Cited by 33
“Following common practice in the field, we use the CLIP zero-shot classifier [Radford et al. 2021] to assess class-level recognition, MS-SSIM [Wang et al. 2003] for image-sketch fidelity following the settings proposed in CLIPascene [2022a], and DreamSim [Fu et al. 2023].”
Image-GS: Content-Adaptive Image Representation via 2D Gaussians
Zhang, Y., Li, B., Kuznetsov, A., Jindal, A., Diolatzis, S., Chen, K., Sochenov, A., Kaplanyan, A., & Sun, Q. (2025) · ACM SIGGRAPH Conference (SIGGRAPH) · doi:10.1145/3721238.3730596 · Cited by 51
“For instance, decreasing the resolution of grid-based methods (ReLU, Instant NGP) leads to block artifacts due to feature vectors being interpolated at sparser locations.”
InfluentialExplicit Semantic-Base-Empowered Communications for 6G Mobile Networks
Wang, F., Zheng, Y., Xu, W., Liang, J., Zhang, P., & Han, Z. (2025) · Engineering · doi:10.1016/j.eng.2025.08.039 · Cited by 16
A ‘Complete Blind’ No-Reference Stereoscopic Image Quality Assessment Algorithm
Appina, B. (2020) · International Conference on Signal Processing and Communications (SPCOM) · doi:10.1109/spcom50965.2020.9179556 · Cited by 15
“SSIM [24], MS-SSIM [25] and VIF [26] are 2D FR IQA models, NIQE [1] is an unsupervised 2D NR IQA model.”
MethodImperceptible Image Steganography Using Symmetry-Adapted Deep Learning Techniques
Khalifa, A. S., & Guzman, A. (2022) · Symmetry · doi:10.3390/sym14071325 · Cited by 16
Neural Face Video Compression using Multiple Views
Volokitin, A., Brugger, S., Benlalah, A., Martín, S., Amberg, B., & Tschannen, M. (2022) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw56347.2022.00182 · Cited by 18
“We report the L1 reconstruction error, PSNR, SSIM and MS-SSIM [19], as well as LPIPS [25].”
MethodTwo-View Mammogram Synthesis from Single-View Data Using Generative Adversarial Networks
Yamazaki, A., & Ishida, T. (2022) · Applied Sciences · doi:10.3390/app122312206 · Cited by 17
“Thereby, we used the peak signal to noise ratio (PSNR), structural similarity (SSIM) [44], multi-scale SSIM (MS-SSIM) [45], and cosine similarity (Cos_sim) to evaluate the similarity between pairs of real and synthesized CC-view mammograms for the test data set described in Section 2.2.”
BackgroundMethodInfluentialLi, Q., Guo, J. J., Fei, Y., Li, F., & Guo, Y. (2022) · SIGGRAPH Asia Conference Papers · doi:10.1145/3550469.3555384 · Cited by 21
Preva: Protecting Inference Privacy through Policy-based Video-frame Transformation
Lu, R., Shi, S., Wang, D., Hu, C., & Zhang, B. (2022) · IEEE/ACM Symposium on Edge Computing (SEC) · doi:10.1109/sec54971.2022.00021 · Cited by 18
“Otherwise, the current frame will be transformed with the saved PrevaPolicy.”
BackgroundMethodInfluentialThe Impact of Excitation Periods on the Outcome of Lock-In Thermography
Sapieta, M., Dekýš, V., Kopas, P., Jakubovičová, L., & Šavrnoch, Z. (2023) · Materials · doi:10.3390/ma16072763 · Cited by 13
Making Video Quality Assessment Models Robust to Bit Depth
Ebenezer, J. P., Shang, Z., Wu, Y., Wei, H., Sethuraman, S., & Bovik, A. C. (2023) · IEEE Signal Processing Letters · doi:10.1109/lsp.2023.3268602 · Cited by 12
“This includes SSIM, MS-SSIM, and VMAF models that are currently globally deployed by streaming and social media platforms.”
MethodSelf-citationMachine learning for cross-scale microscopy of viruses
Petkidis, A., Andriasyan, V., & Greber, U. F. (2023) · Cell Reports Methods · doi:10.1016/j.crmeth.2023.100557 · Cited by 16
“Similarly, the perceived quality of the reconstructed image can be assessed by metrics, such as the multi-scale structural similarity index measure, 134 which can be incorporated into the loss function during network training.”
MethodConfusion Region Mining for Crowd Counting
Zhu, J., Zhao, W., Yao, L., He, Y., Hu, M., Zhang, X., Wang, S., Li, T., & Lu, H. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3311020 · Cited by 14
“…error (mse) loss (4) (5) where e ( x , y ) computes the regress error using a hinge for the background pixels In the training of GEM, we use the hybrid loss of hinged mse and structural similarity (SSIM) [47], which enhance the model’s perceptual ability at the pixel and patch level, respectively.”
Learned Image Transmission over MIMO Fading Channels
Yao, S., Wang, S., Dai, J., & Niu, K. (2023) · IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) · doi:10.1109/pimrc56721.2023.10293784 · Cited by 13
Wang, H., Shu, C., Li, X., et al. (2024) · IEEE Access · doi:10.1109/access.2024.3364050 · Cited by 12
“The objective metrics selected are entropy (EN) [54], standard deviation (SD) [55], edge intensity (EI) [56], visual information fidelity (VIF) [57], Chen-Blum metric ( Q CB ) [58] and MS-SSIM [59].”
BackgroundReducing High-Frequency Artifacts for Generative Model Watermarking via Wavelet Transform
Liu, Y., Li, Z., Wu, H., Wang, Z., & Zhang, X. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3363613 · Cited by 12
“…TITAN RTX 3090 GPU accelerated with CuDNN, which provides acceleration for DNNs. Evaluation Metrics: Four common metrics are applied to assess the quality of the generated images, namely, peak signal to noise ratio (PSNR) [43], structural similarity (SSIM) [43], and multiscale SSIM (MS-SSIM) [44].”
MethodSelf-citationA Dual-Mode Intensity and Polarized Imaging System for Assisting Autonomous Driving
Xie, J. J., Dou, J., Zhong, L., Di, J., & Qin, Y. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3366582 · Cited by 18
“In our case, we adopt the multiscale weighted SSIM (MWSSIM) [36] to establish the loss between input and fused images.”
MethodCGF-Unet: Semantic Segmentation of Sidescan Sonar Based on Unet Combined With Global Features
Sun, Y., Zheng, H., Zhang, G., Ren, J., & Shu, G. (2024) · IEEE Journal of Oceanic Engineering · doi:10.1109/joe.2024.3364670 · Cited by 13
“It can be expressed as [30] where M is the total number of the scales, and p = { p j : j = 1 , . . . , N 2 } and g = { g j : j = 1 , . . . , N 2 } . μ p and μ g are the means of p and g , σ p and σ g are the standard deviations of p and g , and σ pg represents their covariance. β m and γ m indicate…”
BackgroundChen, Z., Sun, H., Zhang, L., & Zhang, F. (2024) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2024.3403524 · Cited by 22
“The overall framework is optimized in an end-to-end way by minimizing the rate-distortion cost, where the distortion term can be calculated with the Mean Squared Error (MSE) or Multi-Scale Structural SIMilarity (MS-SSIM) [139] for optimized objective quality during training.”
MethodEnhancing Ground-Penetrating Radar (GPR) Data Resolution Through Weakly Supervised Learning
Liu, D., Zhou, M., Wang, X., et al. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3410184 · Cited by 12
“The MS-SSIM [54] is employed herein, assessing the internal structure, contrast, and luminance between the original and generated data.”
MethodCIG-STF: Change Information Guided Spatiotemporal Fusion for Remote Sensing Images
You, M., Meng, X., Liu, Q., et al. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3412118 · Cited by 12
“First, we utilize the mean squared error (MSE) and multiscale structural similarity index (MS-SSIM) [49] to calculate the feature loss between the spatiotemporal fusion results and the ground truth.”
MethodLei, D., Zhu, Q., Li, Y., et al. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3416152 · Cited by 12
LLIC: Large Receptive Field Transform Coding With Adaptive Weights for Learned Image Compression
Jiang, W., Ning, P., Yang, J., Zhai, Y., Gao, F., & Wang, R. (2024) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2024.3416831 · Cited by 34
“1) Testing GPU Memory Consumption Comparison: Compared with existing recent learned image compression models Xie’21 [22], Entroformer [70], and NeuralSyntax [71], our proposed LLIC transform coding consumes much less GPU memory.”
Multi-Scale and Multi-Layer Lattice Transformer for Underwater Image Enhancement
Hsu, W., & Hsu, Y.-Y. (2024) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3688802 · Cited by 13
“L1 loss is the difference between the enhancement image (cid:98) 𝑿 and the corresponding reference image 𝑋 , as expressed as follows: where 𝑖 and 𝑗 are the indexes of pixel positions, and 𝐻 and 𝑊 represent the image size.”
Perception-Guided Quality Metric of 3D Point Clouds Using Hybrid Strategy
Zhang, Y.-J., Yang, Q., Xu, Y.-L., & Liu, S. (2024) · IEEE Transactions on Image Processing · doi:10.1109/tip.2024.3468893 · Cited by 12
Saliency-Guided No-Reference Omnidirectional Image Quality Assessment via Scene Content Perceiving
Zhang, Y., Wan, L., Liu, D., Zhou, X., An, P., & Shan, C. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3485447 · Cited by 23
“Wang et al. [41] proposed a multiscale structural similarity method (called MS-SSIM) which supplied more flexibility than previous single-scale methods.”
MethodOptimizing Multi-User Semantic Communication via Transfer Learning and Knowledge Distillation
Nguyen, L. X., Kim, K. T., Tun, Y. L., et al. (2024) · IEEE Communications Letters · doi:10.1109/lcomm.2024.3499956 · Cited by 12
“Due to the page limit, the equation for MS-SSIM can be referred to in [14].”
Semantic Base Enabled Image Transmission With Fine-Grained HARQ
Zheng, Y., Wang, F.-Y., Xu, W.-J., & Zhang, P. (2025) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2025.3532501 · Cited by 17
“Moreover, multi-scale structural similarity (MS-SSIM) [49] and LPIPS are used to measure the perceptual quality of image reconstruction.”
USWformer: Efficient Sparse Wavelet Transformer for Underwater Image Enhancement
Mishra, P., Mehta, N., Vipparthi, S. K., & Murala, S. (2025) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61041.2025.00333 · Cited by 30
Imai, T., Mizutani, T., Iguchi, T., & Haneda, T. (2025) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2025.3568384 · Cited by 18
“The loss on permittivity is comprised of two components: multiscale structural similarity (MS-SSIM) [43] and mean absolute error (MAE) where P i is the ground truth permittivity, and ˆ P i = Y 1 ( F ( x s i )) is the predicted permittivity.”
Jha, S. K., Gupta, V., Sharma, P. J., Mishra, A., & Joshi, S. (2025) · Frontiers in Climate · doi:10.3389/fclim.2025.1572428 · Cited by 28
“Unlike standard SSIM, MS-SSIM is more robust to local variations and noise, providing a more comprehensive assessment of model performance (Wang et al., 2003).”
Robust virtual staining of landmark organelles with Cytoland
Liu, Z., Hirata-Miyasaki, E., Pradeep, S., Rahm, J. V., Foley, C., Chandler, T., Ivanov, I. E., Woosley, H. O., Lee, S.-C., Khadka, S., et al. (2025) · Nature Machine Intelligence · doi:10.1038/s42256-025-01046-2 · Cited by 29
“5DMS − SSIM is the multi-scale structural similarity index 54 measured without downsampling along the depth dimension, and ℒ 𝓁𝓁 1 is the L1 distance (mean absolute error).”
Joint Source-Channel Adaptation for Perceptually Optimized Scalable Video Transmission
Khalek, A. A., Caramanis, C., & Heath, R. W. (2011) · Globecom · doi:10.1109/glocom.2011.6134145 · Cited by 14
“In this section, we describe the proposed system model including the scalable video coding model, the channel model, and the PHY layer model.”
MethodVideo quality measurement based on 3-D. Singular value decomposition
Azar, F. T., Imani, H., & Fathollahian, H. (2014) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2014.12.004 · Cited by 14
“These methods are PSNR, SSIM [3], MS-SSIM [7], Speed SSIM [2], VSNR [8], VQM [15], V-VIF [12], MOVIE [14] and finally latest one, STAQ [16].”
MethodInfluentialDCT-based objective quality assessment metric of 2D/3D image
Liu, X., Sun, C., & Yang, L. T. (2013) · Multimedia Tools and Applications · doi:10.1007/s11042-013-1698-z · Cited by 12
“In addition, Structural Similarity (SSIM) [30] and Multi-scale Structural Similarity Index (MSSIM) [31] proposed by Wang et al. have also been widely accepted, since both algorithms count the structural information instead of error sensitivity-based measurement for quality assessment since the main function of the HVS…”
MethodResultInfluentialLow-complexity enhanced lapped transform for image coding in JPEG XR / HD photo
Maalouf, A., & Larabi, M. (2009) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2009.5413933 · Cited by 12
EVQA: An ensemble-learning-based video quality assessment index
Lin, J. Y., Wu, C.-H., Katsavounidis, I. X., Li, Z., Aaron, A., & Kuo, C. ‐. J. (2015) · IEEE International Conference on Multimedia & Expo Workshops (ICMEW) · doi:10.1109/icmew.2015.7169760 · Cited by 12
“[17] Zhou Wang, Eero P Simoncelli, and Alan C Bovik, “Multiscale structural similarity for image quality assessment,” in Signals, Systems and Computers, 2004.”
BackgroundMethodObjective quality assessment of tone-mapped videos
Yeganeh, H., Wang, S., Zeng, K., Eisapour, M., & Wang, Z. (2016) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2016.7532487 · Cited by 13
Self-citationAn Remapping Operation without Tone Mapping Parameters for HDR Images
Kinoshita, Y., Shiota, S., Iwahashi, M., et al. (2016) · IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences · doi:10.1587/transfun.e99.a.1955 · Cited by 10
The survey of subjective and objective methods for quality assessment of 2D and 3D images
Opozda, S., & Sochan, A. (2014) · Theoretical and Applied Informatics · Cited by 12
Evaluating Multiexposure Fusion Using Image Information
Rahman, H., Soundararajan, R., & Babu, R. V. (2017) · IEEE Signal Processing Letters · doi:10.1109/lsp.2017.2752233 · Cited by 14
Flame temperature estimation from light field image processing
Li, T., Zhang, C., Yuan, Y., et al. (2018) · Applied Optics · doi:10.1364/ao.57.007259 · Cited by 12
Skarbek, W. (2019) · Fundamenta Informaticae · doi:10.3233/fi-2019-1827 · Cited by 13
“where for the signal with b bits per sample the regularizing constants Ci = Ki ·2b, i = 1, 2 (authors of [25] verified K1 = 0.”
BackgroundRobust Image Reconstruction for Block-Based Compressed Sensing Using a Binary Measurement Matrix
Akbari, A., & Trocan, M. (2018) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2018.8451216 · Cited by 12
“It should be noted that the BCS-SPL algorithm works perfectly when a Gaussian random matrix is used for CS. Table 1 and 2 show the quality of reconstructed images at different subrates in terms of PSNR and MSSIM, respectively.”
MethodInfluentialWang, X., Shao, F., Jiang, Q., et al. (2019) · IEEE Access · doi:10.1109/access.2019.2891070 · Cited by 11
Perceptually-inspired super-resolution of compressed videos
Ma, D., Afonso, M. F., Zhang, F., & Bull, D. (2019) · SPIE Applications of Digital Image Processing · doi:10.1117/12.2530688 · Cited by 20
Practical Stacked Non-local Attention Modules for Image Compression
Liu, H., Chen, T., Shen, Q., & Ma, Z. (2019) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · Cited by 16
Okarma, K., & Fastowicz, J. (2019) · Elektronika ir Elektrotechnika · doi:10.5755/j01.eie.25.5.24357 · Cited by 15
Gan-Based Effective Bit Depth Adaptation for Perceptual Video Compression
Ma, D., Zhang, F., & Bull, D. (2020) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme46284.2020.9102865 · Cited by 14
“The training process is similar to that in [26], in which the BDGAN generator (BDNet) was firstly trained using the multi-scale structural similarity index (MS-SSIM) quality metric [37] as the loss function.”
Method
Metadata
- Authors
- Wang, Z., Simoncelli, E. P., & Bovik, A. C.
- Affiliations
- Courant Institute of Mathematical Sciences (US)
- The University of Texas at Austin (US)
- Venue
- The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003
- Publisher
- IEEE
- Published
- 2004-07-08
- Pages
- 1398-1402
- ISSN
- 1058-6393 (ISSN-L)
- OpenAlex
- W1580389772
- Cited by
- 6,253
- References
- 15
- Citations by year
- 2026: 344 · 2025: 675 · 2024: 729 · 2023: 668 · 2022: 662 · 2021: 641 · 2020: 563 · 2019: 438 · 2018: 305 · 2017: 267 · 2016: 206 · 2015: 173 · 2014: 166 · 2013: 112 · 2012: 89
- FWCI
- 3.98
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Asilomar Conference on Signals, Systems and Computers (ACSSC) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image and Video Quality Assessment
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image Retrieval and Classification Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Keywords
- structural similarity, image quality assessment, multiscale analysis, human visual system, perceptual image quality, image generation, parameter calibration, viewing conditions
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.