ABBYY Mobile OCR Engine

Step 1: Image Import and Processing

The image is loaded from memory and prepared for OCR. Image binarisation separates text from the background, producing a black-and-white image that is much smaller in size than the colour original. Additional skew correction and document orientation detection can be applied.

Automatic image skew correction function. When you take photographs with a mobile device camera or scan images with a portable scanner on-the-run, image skews occur fairly often, which has a negative impact on the recognition quality. ABBYY Mobile OCR Engine allows detection and correction of skews within one degree of precision, which results in a significant improvement of the quality and accuracy of mobile OCR. 

Document orientation detection function. Image pre-processing automatically detects the orientation of a page of text to be recognized (if it is sideways or upside down).

Step 2: Document Analysis

Document Analysis is a set of algorithms that analyses the image — it detects letters, joins the letters into words, then into lines of text, and finally, into paragraphs. Additionally, the reading area is cleaned and noise removed.

Hyphenation support. If the engine encounters a part of a hyphenated word (e.g. Mon-) on one line and the second part (e.g. day) on the next line,  it will join them into one (Monday).

Preserving multi-column text. In the previous versions of ABBYY Mobile OCR Engine, text was recognized left to right top to bottom strictly, which resulted in placing all the recognized data into one linear massive. ABBYY Mobile OCR Engine 4.0 has a function called «Paragraph Assembly», thanks to which the new DA identifies text block borders and recognizes each block separately, recognizing text left to right and top to bottom only within a separate block, thus preserving the format of a multi-column text, paragraphs, and text segmenting.

Preserving Character Fonts. ABBYY Mobile OCR Engine identifies the font properties of a source text, i.e. «bold type», «italic» or «underlined».

Confidence level indicator. This function shows the level of certainty for recognized text, allowing developers to set flexible criteria for implementation of proofreading and verification functions.   

Spell checking during text recognition also considerably improves the quality of the output text

Step 3: Optical Character Recognition (OCR)

Then the detected blocks on the image are recognized using the special language and pattern definitions. If dictionaries are available, then the texts are also compared to improve the overall recognition quality. Recognition results are the set of characters with coordinates united in lines. Each character has the level confidence which show how recognition engine was sure in final character choice.

Two mobile recognition modes:

• Fast mode of express recognition – it is most convenient when the image is of good quality and allows cutting down the time required for its recognition and processing.

• Full mode for accurate recognition – it is best for low-quality images, when more time is required to achieve the optimal result.

Business Card Processing (optional)

The recognition results are analyzed and the relevant contact information from business cards is extracted.

This function allows retrieving information from business card images, such as first name, last name, position of the cardholder, various types of phone numbers, e-mail, company name, Web site and postal address of the company.

Now Business Card Reading technology allows to recognize business cards in 23 languages.

Barcode recognition (optional)

This function provides barcode recognition that supports many types of 1D and 2D barcodes.Searching for a barcode on a page and barcode detection are not available. The user should either capture only the barcode or manually crop the image to the barcode.  

Step 4: Result Processing

The recognition results can be processed and exported. The developer of the application has full control over the OCR results.