OCR

An encyclopedic cluster explaining optical character recognition: how the recognition pipeline turns scanned page images into machine-readable text, what governs accuracy, how searchable output and text layers are produced, and how languages, scripts, and layout are handled. Standards-first and vendor-neutral, covering OCR output formats (searchable PDF text layers, hOCR, ALTO XML) and adjacent recognition types (ICR, OMR) without benchmarks, rankings, or invented accuracy figures.

34 live pages · long-term capacity 26–42

Entities

technology

Optical Character Recognition · Tesseract

concept

Intelligent Character Recognition · Optical Mark Recognition · Text layer · Binarization · Deskew · Layout analysis · Ground truth · Confidence score · Handwriting recognition

format

Searchable PDF · hOCR · ALTO XML · Mixed Raster Content

standard

PDF/A · Unicode · OCR-A

organization

Adobe · ISO

Connected clusters

In the archive

Planned coverage

How OCR Works: From Page Image to Text — Walks the recognition pipeline: image capture, preprocessing, layout analysis, character/word recognition, and text output.
What Affects OCR Accuracy — Explains the durable factors — resolution, contrast, skew, noise, font, language, layout complexity — that influence recognition quality, without citing accuracy percentages.
How Searchable PDFs Are Created — Describes how OCR adds an invisible text layer over a scanned page image so the document becomes selectable and searchable.
OCR vs ICR: Printed Text vs Handwriting Recognition — Contrasts optical character recognition of machine print with intelligent character recognition aimed at hand-printed characters.
How OCR Handles Languages and Scripts — Explains language models, character sets, and script direction (Latin, CJK, right-to-left) and why script selection matters.
Preparing Documents for OCR — Practical, vendor-neutral scan-quality guidance: resolution, lighting, contrast, and flat pages that improve recognition.
Image Preprocessing for OCR: Deskew, Binarize, Denoise — Explains the preprocessing steps that clean a page image before recognition and why each helps.
OCR Confidence Scores Explained — Describes what per-character and per-word confidence values represent and how they flag uncertain recognition.
How OCR Reads Tables and Multi-Column Pages — Covers layout/zoning analysis and reading order for complex page structures.
Why OCR Makes Mistakes — Explains common error sources — look-alike characters, broken glyphs, bleed-through, unusual fonts — as durable phenomena.
OCR Output Formats: Text, Searchable PDF, hOCR, ALTO — Compares plain text, searchable PDF, hOCR, and ALTO XML and what positional/layout data each preserves.
OCR and Handwriting Recognition — Explains why cursive and free handwriting are harder than machine print and how handwriting recognition differs.
Searchable PDF — Defines a searchable PDF as an image page plus a hidden text layer.
Text Layer — Defines the invisible OCR text layer positioned over a page image.
hOCR — Defines the hOCR format for embedding OCR results, including layout, in HTML/XML.
ALTO XML — Defines ALTO XML as a standard for representing OCR text and layout of digitized pages.
Binarization — Defines converting a grayscale/color scan into a two-tone image for recognition.
Deskew — Defines correcting the rotation of a tilted scanned page.
Intelligent Character Recognition (ICR) — Defines ICR as recognition aimed at hand-printed characters.
Optical Mark Recognition (OMR) — Defines OMR as detecting marks in fixed positions, e.g. checkboxes and bubble forms.
Ground Truth — Defines verified reference text used to evaluate or train OCR.
Batch OCR of Scanned Archives — Vendor-neutral workflow for running OCR across many scanned files to make an archive searchable.
Make Old Scans Searchable with OCR — Workflow for adding a text layer to existing image-only scans.
Extract Text from a Photo of a Document — Workflow for OCR on phone-camera captures, including cropping, deskew, and contrast steps.