Recognition settings

Aspose.OCR for JavaScript via C++ provides good recognition accuracy and performance out of the box. Nevertheless, there will inevitably be cases where the default settings fail to deliver dependable recognition results.

To fine-tune the recognition settings, pass an optional Module.WasmAsposeOCRRecognitionSettings object to the recognition function.

Language configuration

Set the recognition language.

Setting	Value	Default value	Description
language_alphabet	Object	Module.Language.NONE	Specify a language for recognition. It is highly recommended that you always provide this setting.
alphabet	String	All symbols	A custom list of characters to be recognized, provided as a case-sensitive string. Characters that do not match the provided list are ignored.
allowed_characters	Object	Module.CharactersAllowedType.ALL	The predefined whitelist of characters Aspose.OCR engine will look for.
ignoredCharacters	String	none	A blacklist of characters that are ignored during recognition.

Document structure detection

Determine how to handle complex text layouts (multi-column texts, tables, and so on).

Setting	Value	Default value	Description
detect_areas_mode	detect_areas_mode_enum	detect_areas_mode_enum::DOCUMENT	Manually override the default document areas detection model.
lines_filtration	Boolean	false	Set to true to recognize text in tables. Set to false to improve performance by ignoring table separators and treating tables as plain text lines.
all_image	Boolean	false	Force recognition of the entire image as a single block of text. It is recommended to enable (set to true) this setting only when working with very simple, one-line images.

Preprocessing

Improve recognition accuracy for low-quality images.

Setting	Value	Default value	Description
correct_skew	Boolean	true	Automatically correct image tilt (deskew) before proceeding to recognition. Since most of the images have some degree of tilt, it is recommended to keep this setting always on.
skew	Floating point number	0.0	Manually rotate the image by the specified degree. Recommended for significantly tilted images (by more than 15° either direction), for which the automatic skew correction may fail. -360 to 0: rotate counterclockwise; 0 to 360: rotate clockwise.
auto_contrast	Boolean	false	Automatically increase the contrast of images before proceeding to recognition. Enable for blurry images and out-of-focus photos.
auto_denoising	Boolean	false	Automatically remove noise from images before proceeding to recognition.
upscale_small_font	Boolean	false	Improve small font recognition and detection of dense lines. Recommended for endnotes, food labels and other images with small text.
threshold_value	Number	0	Override the automatic binarization settings.

Example

The following code example shows how to fine-tune recognition:

// Prepare images
var source = Module.WasmAsposeOCRInput();
source.url = filename;
var content = new Module.WasmAsposeOCRInputs();
content.push_back(source);
// Recognize rotated photos of book pages in Ukrainian
var settings = Module.WasmAsposeOCRRecognitionSettings();
settings.language_alphabet = Module.Language.UKR;
settings.skew = -90;
settings.detect_areas_mode = Module.DetectAreasMode.CURVED_TEXT;
// Recognize image
var result = Module.AsposeOCRRecognize(content, settings);