Adaptive Quality Mask Harvesting for Pixel-Wise Quality-Weighted Deep-Sky Reconstruction

Adaptive Quality Map Hyperstacking (AQMH) v0.2.0 describes a deterministic pixel-wise quality-weighted reconstruction method for short-exposure deep-sky imaging. The method replaces block-constant local tile weights with dense per-frame quality maps and reconstructs each supported output pixel from registered frame samples using AQMH weights evaluated at the same pixel. This Zenodo bundle contains the AQMH method paper, diagnostic figures, and compact summary data for an M31 validation run. The run processed 645 frames on a 3924 x 2310 canvas, generated 645 half-resolution luminance quality maps, reconstructed without Classic Tile Compile fallback, and reported improved run-level FWHM and background RMS diagnostics. The v0.2.0 release adds a background-gradient penalty in global frame quality, a registration-weight guard, adaptive low-frequency neutralisation, and structure-masked detail blending — all gated by a uniform-control validation that preserves the raw AQMH output when no post-processed candidate passes all regression thresholds. The release documents the AQMH v0.2.0 method definition, implementation constraints, reconstruction invariants, post-reconstruction robustness extensions, quality-map diagnostics, and limitations. It is intended as a method and implementation validation artifact, not as a full statistical benchmark across datasets or instruments.