Use exact controls and repeated parser-valid recovery in PDF417 preprocessing studies

When replaying blurred PDF417 crops, keep the preserved exact original as the control and test a small deterministic ladder of non-generative transforms. Vary one preprocessing family at a time, use the same decoder and parser, and repeat both control and candidate decodes. Judge an apparent rescue by stable parser-valid recovery and, when authorized ground truth exists, payload consistency rather than visual sharpness alone. Do not publish decoded payloads or private image content.

Crop-only replay is a prioritization experiment, not a faithful test of a realtime guide lane. Deployed-path validation requires the matching full frame, crop-generation path, encoding, decoder configuration, and repeated outcomes. An isolated recovery does not establish identity correctness, another decoder's behavior, realtime performance, or production benefit.