OCR and Image processing
What image formats are supported?
- ABBYY Mobile OCR SDK functions work with images in FINE_IMAGE format. You must load image into the memory and convert it to FINE_IMAGE format before calling the functions.
- ABBYY Mobile OCR SDK distribution kit includes sample with the procedure of loading only BMP image files. If you want to load other image files, you should create your own procedure of loading and converting.
- FINE_IMAGE format is very similar to BMP format and it can be easily obtained from standard image representation used on a certain platform (e.g. from HBITMAP on WinMobile, from CFbsBitmap on Symbian, from UIImage on iPhone, etc).
- It is also important to specify appropriate resolution in dpi (Resolution field of FINE_IMAGE structure) otherwise text on image can be lost. In case you are not sure what resolution to specify, set 0 for business cards.
Does ABBYY Mobile OCR Engine support CJK languages?
- Up to Version 2 there was only a special CJK library for WinMobile 6 platform
- Since Version 4 of Mobile OCR Engine Chinese Japanese and Korean OCR is available for all supported operating systems
Does ABBYY Mobile OCR Engine support automatic orientation detection?
If no, when will it be implemented?
- ABBYY Mobile OCR Engine 3.0 doesn’t support this feature.
- Automatic orientation detection is planned to be implemented in version 4.0, but not in the first release (February/March 2010)- it is planned for a maintenance release.
Does ABBYY Mobile OCR Engine support skew correction?
What is the maximum angle which can be detected?
- MSDK 2.0 corrects skew up to 5 degrees.
- ABBYY Mobile OCR Engine 3.0 corrects skew up to 10-15 degrees.
Can I create a user-defined language in ABBYY Mobile OCR Engine?
- ABBYY Mobile OCR Engine itself doesn’t support creation of user language. However, it can be created in FineReader 4.0 and then can be used in ABBYY Mobile OCR Engine. To get FineReader 4.0 distribution please contact ABBYY. Description of how to create user language can be found in FineReader 4.0 documentation.
- In order to use the created language in the Mobile SDK, you need to transform language created in FineReader to ABBYY Mobile OCR Engine pattern file. This procedure is described in the MSDK Help file - “How to Create a Pattern File” topic (“Programmer's Guide”→“How to…?”).
Can I use pattern training in ABBYY Mobile OCR Engine?
No, ABBYY Mobile OCR Engine doesn’t support pattern training.
Does ABBYY Mobile OCR Engine support photography distortion correction?
Optical/ photography distortions are:
- correction of “fish eye”
- bow-shaped distortion
- trapezium distortion
No, current version of ABBYY Mobile OCR Engine doesn't support this functionality.
Why do I receive “TRIAL” or “ABBYY” words in recognition results?
This means that you are running your application with a trial license key.
What are image requirements?
In general, the main requirements for images are the following:
- The text on image must be sharp and have enough contrast
- The camera should have auto-focus and min. 3 mega pixel
- When image is acquired by camera, the text has to be in focus
- The image should have no photography distortions e.g. fisheye lens effect; trapezium distortions
Are there any recommendations for taking photos?
Please refer to photographing documents instruction.
How can I improve recognition quality?
ABBYY Mobile OCR Engines are based on adjusted OCR technologies of FineReader 4.0, and in most cases recognition quality is limited by OCR technology and there is no possibility to improve it significantly. The general recommendation is to improve source image quality. In some cases using a user-defined recognition language can help to improve OCR quality to some extent.
