Technical Report v1.20: A Ground-Truth-Free Binary Classifier for SSIM Amplification Phase Risk, Validated Across Four Models

This series has established mu_stat as a model-specific, empirically-calibrated transition point marking SSIM's Amplification Phase -- the region where the luminance term becomes hypersensitive to small absolute deviations between ground-truth (GT) mean brightness mu_gt and a reconstruction's own mean brightness mu_notta (Technical Reports v1.10, v1.14). Diagnosing whether a given block falls inside this region has so far required mu_gt itself, i.e. GT. This report tests whether a practitioner with only a model's reconstruction, no GT, can still obtain a useful Amplification Phase risk signal using mu_notta alone. Exact value estimation (mu_gt recovered from mu_notta) proves structurally unreliable: the best method tried across a wide range of regression, lookup-table, and hybrid variants improves at most ~65% of individual blocks, with a persistently heavy-tailed error distribution, a consequence of mu_gt's underlying 8-bit quantization making its finest levels difficult to separate using mu_notta alone. Reframing the question as binary classification instead -- is this block at risk, yes or no -- sidesteps that identifiability problem entirely, since mu_stat falls partway through the quantization grid and every finer distinction below it is irrelevant to the binary question. A single untrained threshold, mu_notta <= mu_stat (the exact same calibrated value, no fitting on mu_notta, with mu_notta itself recovered algebraically from a stored SSIM component), classifies Amplification Phase risk at 96-100% accuracy across all four models in this series, with recall at or above 99.5% throughout. Because mu_stat itself is calibrated from GT only once per model, this shifts when GT is needed -- from every image checked to once per model -- rather than eliminating it. The contribution is offered as a validated diagnostic method for a specific, narrow use case -- flagging likely Amplification-Phase-risk blocks without per-image GT -- not as a claim that GT-free SSIM diagnosis is possible in general. Every finding is computed against the same four independent, no-join, from-scratch verification databases (Technical Report v1.12), read-only, screen-recorded and hash-verified. The complete verification SQL, covering all four models sequentially, is included to enable independent reproduction. Dataset and Mandatory Citation: - Source: NIH ChestX-ray8 (Hospital-scale chest x-ray database) - Citation: Wang, X., Peng, Y., Lu, L., Lu, Z., Bagheri, M., & Summers, R. M. (2017). "ChestX-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thoracic diseases." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 3462–3471. - Download: https://nihcc.app.box.com/v/ChestXray-NIHCC Contact: [email protected] (individual) /[email protected] (Meritcs) Published by Meritcs (https://merit-metrics.org), an independentresearch organization studying image-quality metric reliability.