What is SSIM?
SSIM — SSIM (Structural Similarity Index Measure) is a perceptual image quality metric that compares two images for luminance, contrast, and structural similarity. Values range from 0 to 1; SSIM 0.95+ is considered visually lossless.
SSIM (Structural Similarity Index Measure) is a perceptual quality metric developed by Wang et al. (2004) to measure the similarity between two images from a human visual system perspective. Unlike PSNR (Peak Signal-to-Noise Ratio), which measures raw pixel differences, SSIM models how humans perceive structural patterns, luminance, and contrast.
SSIM computes three components — luminance similarity, contrast similarity, and structural correlation — across sliding windows of the image, then combines them into a single score between 0 (no similarity) and 1 (identical). A score of 0.95+ is generally considered visually lossless; scores below 0.85 typically show perceptible quality degradation.
SSIM is widely used to evaluate image codec performance, compare compression algorithms, and set quality targets for web image optimization. Codec benchmarks (Google's WebP tests, Netflix's VMAF research) use SSIM alongside PSNR and VMAF to measure how closely a compressed image matches the original.
SSIM vs PSNR: Which to Use?
PSNR (Peak Signal-to-Noise Ratio) measures the ratio of maximum possible signal power to the noise power introduced by compression. It is simple to compute but correlates poorly with human perception — an image can have high PSNR but visible artifacts. SSIM correlates better with human perception of quality because it models structural patterns rather than raw pixel differences. For quality assessment of compressed images, SSIM is the preferred metric over PSNR.
SSIM in Practice
Image optimization tools use SSIM targets to find the minimum quality setting that keeps perceived quality above a threshold. The Butteraugli perceptual metric and MS-SSIM (multi-scale SSIM) are further refinements used in advanced codec evaluations. For day-to-day web image optimization, aiming for SSIM > 0.90 at the compressed quality setting ensures the output will be indistinguishable from the original to most viewers.
SSIM Tools on PicsSizer
Frequently Asked Questions
What is a good SSIM score for compressed images?▾
SSIM > 0.95 is considered visually lossless — most viewers cannot detect the compression. SSIM 0.90–0.95 has minor artifacts visible only on close inspection. SSIM 0.80–0.90 shows noticeable compression artifacts. Below 0.80, significant quality loss is evident.
How is SSIM different from visual quality ratings?▾
SSIM is an objective mathematical metric computed from pixel values. Visual quality ratings (MOS — Mean Opinion Score) are subjective scores from human raters. SSIM correlates well with MOS for most image types but is not perfect — some image manipulations score high on SSIM but are perceptually disturbing (e.g., phase-shifted sine waves).