Phase 4 turned stored OCR blocks into a retrieval system. Exact, case-insensitive, and literal phrase matching produced contextual snippets, while a deterministic ranker used match quality, confidence, position, count, recency, and asset state.
Phase 5 completed the visual connection. OCR bounding boxes use normalized source-image coordinates, and geometry helpers validate, convert, and map those regions into the aspect-fit image shown on screen.
The detail view can distinguish ordinary, matched, selected, and selected-matched regions. Search therefore no longer ends at a text row; it returns to the original screenshot and the place where the recognized words appeared.
Basic zoom, recognized-text selection, copy actions, and early accessibility labels made that connection usable without turning SnipDig into an image editor.
← Back to SnipDig