The 8 fields and concepts you'll meet in the response — defined in plain English, each with a real example value.
Optical Character Recognition—technology that converts images of text into machine-readable text.
OCR analyzes image pixels to identify text characters. Modern OCR uses AI/deep learning for high accuracy across fonts, languages, and conditions. Applications include document digitization, receipt scanning, and accessibility tools.
ExampleConverting a scanned document into editable text
Dots Per Inch—a measure of image resolution, particularly for scanned documents.
300 DPI is the standard for document scanning, providing enough detail for accurate OCR. Lower DPI loses fine character details. Higher DPI increases file size without significantly improving OCR accuracy. DPI relates to print resolution; for screen images, pixel dimensions matter more.
Example300 DPI scan of a document, 72 DPI screen capture
The OCR stage that locates regions of an image containing text.
Before recognizing characters, OCR must find where text appears. Text detection identifies bounding boxes around text regions, handles different orientations, and separates text from images, backgrounds, and noise. Modern detectors use neural networks.
ExampleFinding the text blocks in a magazine page with photos
The OCR stage that identifies individual characters from image segments.
After detecting text regions, OCR segments and classifies each character. Traditional methods match against templates; modern approaches use neural networks that learned from millions of character examples. Confidence scores indicate certainty.
ExampleIdentifying "A" from a segment of a document image
Image enhancements applied before OCR to improve recognition accuracy.
Preprocessing steps include noise reduction, contrast enhancement, binarization (converting to black/white), deskewing (straightening rotation), and background removal. Good preprocessing significantly improves OCR accuracy on low-quality images.
ExampleEnhancing contrast and deskewing a tilted photo of a receipt
Converting an image to pure black and white (no grayscale) for OCR processing.
Binarization separates text (black) from background (white) using a threshold. Adaptive thresholding handles uneven lighting by varying the threshold across the image. Proper binarization improves character segmentation and recognition.
ExampleConverting a grayscale scan to crisp black text on white background
A numerical measure of how certain the OCR system is about its text extraction.
Confidence scores range from 0 to 1 (or 0-100%). Higher scores indicate more certain recognition. Low confidence may mean poor image quality, unusual fonts, or ambiguous characters. Applications can flag low-confidence results for human review.
Example0.95 confidence = 95% certain of the extracted text
A writing system used for a language, such as Latin, Cyrillic, Arabic, or Chinese.
Different scripts require different OCR approaches. Latin (English, Spanish), Cyrillic (Russian), Greek, Arabic (right-to-left), Hebrew (right-to-left), and CJK (Chinese, Japanese, Korean—thousands of characters) each have unique recognition challenges. Modern OCR supports multiple scripts.
ExampleLatin script for English, Devanagari script for Hindi
