ssim

**SSIM** is the **Structural Similarity Index metric that compares luminance, contrast, and structure between two images to estimate perceived similarity** - it is widely used for evaluating restoration and compression quality. **What Is SSIM?** - **Definition**: Full-reference image metric designed to correlate better with perception than pixel error alone. - **Core Components**: Combines local comparisons of brightness, contrast, and structural patterns. - **Score Range**: Typically reported from 0 to 1 where higher values indicate stronger similarity. - **Evaluation Scope**: Commonly applied in denoising, super-resolution, compression, and enhancement studies. **Why SSIM Matters** - **Perceptual Relevance**: Captures structural distortions that PSNR can miss. - **Benchmark Adoption**: Standard metric in image-processing papers and production QA pipelines. - **Model Tuning**: Useful for selecting checkpoints that preserve scene structure. - **Regression Detection**: Highlights quality drops after codec or model updates. - **Interpretability**: Component-wise structure view helps diagnose artifact type. **How It Is Used in Practice** - **Window Configuration**: Use consistent patch size and boundary handling for fair comparison. - **Metric Pairing**: Combine SSIM with PSNR and perceptual metrics for balanced evaluation. - **Dataset Coverage**: Evaluate across textures, edges, and low-light scenes to avoid bias. SSIM is **a core structural-fidelity metric in image-quality evaluation** - SSIM is most useful when reported with complementary perceptual and distortion measures.