[ ]

When photographing pages of books and magazine articles, the cylindrical curvature of the page results in distortion of the image, causing the lines of text to curl. This 3D geometric distortion negatively affects recognition.

DetectAreasMode.CURVED_TEXT areas detection mode uses a specialized neural network that automatically tracks and rectifies curved lines of text. This greatly improves recognition accuracy and allows much more text to be recovered and extracted.

Detecting and rectifying curved lines of text

However, this algorithm is less efficient and consumes more resources when dealing with perfectly straight images, such as a single scanned sheet of paper. Try DetectAreasMode.DOCUMENT or DetectAreasMode.PHOTO instead.


The following code sample demonstrates how to use this document areas detection algorithm:

Aspose.OCR.AsposeOcr recognitionEngine = new Aspose.OCR.AsposeOcr();
// Add an image to OcrInput object
Aspose.OCR.OcrInput input = new Aspose.OCR.OcrInput(Aspose.OCR.InputType.SingleImage);
// Set document areas detection mode
Aspose.OCR.RecognitionSettings recognitionSettings = new Aspose.OCR.RecognitionSettings();
recognitionSettings.DetectAreasMode = Aspose.OCR.DetectAreasMode.CURVED_TEXT;
// Recognize image
List<Aspose.OCR.RecognitionResult> results = recognitionEngine.Recognize(input, recognitionSettings);
foreach(Aspose.OCR.RecognitionResult result in results)