How to Read Images in C# with IronOCR

Why does confidence level matter?

The confidence level indicates IronOCR's certainty about extracted text accuracy. Values above 85% generally indicate reliable results. Lower scores may require image preprocessing or manual review. Use confidence scores to automatically flag documents for human verification or trigger additional image optimization filters.

When should I use different image formats?

PNG and TIFF formats provide the best OCR results due to lossless compression. Use PNG for single-page documents and TIFF for multi-page scans. JPEG works well for photographs but may introduce compression artifacts. BMP offers uncompressed quality but larger file sizes. GIF suits simple graphics with limited colors. Learn more about format-specific optimization.

What are common image reading errors?

Common errors include low image resolution (below 200 DPI), skewed text, poor contrast, or unsupported languages. IronOCR provides automatic correction for many issues, but severe problems may require manual preprocessing. See our troubleshooting guide for solutions.

When should I use byte arrays over file paths?

Byte arrays work best when images come from databases, web services, or encrypted sources. They provide better security since files don't need temporary disk storage. Use byte arrays for cloud applications, microservices, or when processing sensitive documents. File paths remain more efficient for local batch processing of large image collections.

using IronOcr;
using IronSoftware.Drawing;
using System.IO;

// Method 1: From URL
var imageFromUrl = AnyBitmap.FromUri("https://example.com/document.jpg");
using var urlInput = new OcrImageInput(imageFromUrl);

// Method 2: From Stream
using var fileStream = File.OpenRead("document.png");
using var streamInput = new OcrImageInput(fileStream);

// Method 3: From System.Drawing (with IronSoftware.Drawing)
var bitmap = AnyBitmap.FromFile("scan.bmp");
using var bitmapInput = new OcrImageInput(bitmap);

// Process any of these inputs
IronTesseract ocr = new IronTesseract();
OcrResult result = ocr.Read(bitmapInput);

using IronOcr;
using IronSoftware.Drawing;
using System.IO;

// Method 1: From URL
var imageFromUrl = AnyBitmap.FromUri("https://example.com/document.jpg");
using var urlInput = new OcrImageInput(imageFromUrl);

// Method 2: From Stream
using var fileStream = File.OpenRead("document.png");
using var streamInput = new OcrImageInput(fileStream);

// Method 3: From System.Drawing (with IronSoftware.Drawing)
var bitmap = AnyBitmap.FromFile("scan.bmp");
using var bitmapInput = new OcrImageInput(bitmap);

// Process any of these inputs
IronTesseract ocr = new IronTesseract();
OcrResult result = ocr.Read(bitmapInput);

IRON VB CONVERTER ERROR developers@ironsoftware.com

Why does memory management matter for image bytes?

Large images consume significant memory, especially when processing multiple documents simultaneously. Using 'using' statements ensures proper resource disposal. For batch processing, consider implementing a queue system with limited concurrent operations. The multithreading guide demonstrates efficient memory management techniques.

What are the performance implications of different input types?

File paths offer the fastest performance for local files as IronOCR reads data directly. Byte arrays require loading entire images into memory but provide flexibility. Streams balance memory usage and performance by reading data incrementally. For optimal performance with large batches, see our performance tuning guide.

Why does specifying regions improve performance?

Processing only relevant image areas reduces computational overhead by 60-90%. OCR engines analyze every pixel in the input area, so smaller regions mean faster processing. This approach also improves accuracy by eliminating potential interference from headers, footers, or decorative elements outside the target text area.

When should I use multiple scan regions?

Use multiple regions for documents with distinct text areas like forms, invoices, or multi-column layouts. Process each region separately to maintain logical text flow. This approach works well for extracting table data or reading specific fields from structured documents.

What are the coordinate system conventions?

IronOCR uses standard pixel coordinates with origin (0,0) at the top-left corner. X increases rightward, Y increases downward. Rectangle parameters are (X, Y, Width, Height). For precise region selection, use image editing tools to identify pixel coordinates or implement a visual region selector in your application.

When should I apply image preprocessing filters?

Apply filters when dealing with scanned documents, photographs of text, or images with quality issues. Common scenarios include fixing skewed pages, removing background noise from photocopies, or enhancing faded text. The DPI settings guide helps optimize resolution-related issues.

Why does filter order matter?

Filter sequence significantly impacts results. Apply rotation correction (Deskew) first, followed by noise removal, then contrast enhancement. Binarization should typically come last. Incorrect ordering can amplify problems – for example, sharpening before denoising increases noise visibility. Test different sequences for optimal results.

What are common preprocessing mistakes?

Over-processing is the most common error. Excessive sharpening creates artifacts, aggressive denoising removes fine text details, and improper binarization thresholds lose information. Start with minimal preprocessing and add filters only when needed. The image quality correction guide provides detailed best practices.

Why does instance reuse improve performance?

IronTesseract initialization loads language data and configures the OCR engine, taking 200-500ms. Reusing instances eliminates this overhead for subsequent operations. Create a singleton instance for web applications or a shared instance for batch processing to maximize efficiency.

When should I use parallel processing?

Parallel processing benefits scenarios with multiple independent images. Process different pages or documents simultaneously, but avoid parallelizing operations on the same image. Modern CPUs handle 4-8 concurrent OCR operations effectively. Monitor memory usage as each operation requires 100-500MB depending on image size.

What are the memory usage considerations?

OCR operations typically require 10-20x the image file size in RAM. A 5MB image may use 50-100MB during processing. For large batches, implement a producer-consumer pattern with limited concurrent operations. The abort token example demonstrates cancellation for memory-intensive operations.