Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

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.

Multiscale structural similarity for image quality assessment

Wang, Z., Simoncelli, E.P., Bovik, A.C. · 2004 · The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003

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

Crossref · 10.1109/acssc.2003.1292216

Cites

16 of 16 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 Valid · 1 Not valid · 14 Not checked · 18 Own results and interpretations

10 more claims

Create a free account to see all 33 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 8
  • Probably not a substantial claim. 13

Create a free account to see these claims.

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?

Citations

Cites

Show all 15 references

Research works printed but not linked: 2

  1. 0, JNDmetrix Technology, Sarnoff Corporation
  2. 2000, Final report from the video quality experts group on the validation of objective models of video quality assessment, VQEG

Cited by

See all 6,253

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Multiscale structural similarity for image quality assessment. https://app.cochat.ai/papers/10.1109/acssc.2003.1292216
The paper
Wang, Z., Simoncelli, E. P., & Bovik, A. C. (2004). Multiscale structural similarity for image quality assessment. The Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003. https://doi.org/10.1109/acssc.2003.1292216

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.