Optical character recognition

Optical character recognition (OCR) is software that detects text in images and converts it into machine-readable characters. It can preserve layout or produce plain text, depending on the system.

Image preprocessing: Image preprocessing transforms an image to improve its suitability for later analysis. Deskewing, denoising, and contrast adjustment can make scanned text easier to recognize.
Raster graphics: Raster graphics represent images as grids of pixels, each storing color or intensity information. Scanned pages and camera images enter OCR as pixel grids.
Searchable PDF: A searchable PDF combines page images with text data that supports searching and selection. OCR supplies the text layer that makes scanned PDFs searchable.
Optical mark recognition: Optical mark recognition detects filled or marked regions, such as bubbles on forms. Unlike OCR, it reads locations or marks rather than identifying arbitrary printed characters.
Text detection: Text detection locates text regions in an image without necessarily identifying their contents. OCR systems first need to find where lines, words, or characters appear.
Binarization: Binarization converts an image into two-valued regions, commonly foreground and background. Separating ink from paper can simplify text segmentation in degraded scans.
Document digitization: Document digitization converts paper records into digital representations and workflows. OCR makes digitized documents usable as text rather than images alone.
Barcode: A barcode encodes data in a visual pattern designed for machine reading. Barcodes encode symbols directly; OCR interprets the appearance of written characters.
Character recognition: Character recognition identifies symbols in visual input and assigns them character labels. This is the stage that turns detected letter shapes into machine-readable characters.
Convolutional neural network: A convolutional neural network is a neural network designed to learn spatial patterns in grid-like data. Convolutional layers can learn visual features that distinguish character shapes.

Linked from 15 pages