What problem does it solve?
Extracting text from images in MATLAB often fails due to poor preprocessing, wrong detection methods, or misconfigured ocr() parameters, producing garbage output and wasted trial-and-error cycles. This Skill provides a structured diagnose-preprocess-detect-recognize-validate workflow that routes each image type to the correct pipeline.
Core Features & Use Cases
- Guided OCR Pipeline: Diagnose the image visually, then apply the prescribed preprocessing route (binarization, deskewing, local contrast subtraction, SAM segmentation) before calling ocr().
- Text Detection Methods: Choose between detectTextCRAFT for natural scenes, MSER for documents, manual ROI for fixed layouts, and imsegsam for text on textured backgrounds.
- Validation and Logging: Compute CER/WER with evaluateOCR against ground truth, then save a re-runnable pipeline script and decision log.
- Use Case: Read stamped serial numbers from metal parts, extract text from scanned receipts, read seven-segment meter displays, or digitize multi-language documents.
Quick Start
Ask your AI agent to extract the text from an image file using MATLAB OCR, and it will diagnose the image, build the pipeline, and save the results.