For using CloudOCR SDK, does ABBYY provide a client side SDK (Android, iOS or Windows) that does binarization of the image so that size of the image is reduced even better than zipping it? If not, are there any suggested algorithms or open source implementations? I would imagine it would offer a 8x or more reduction in size compared to greyscale or color. In addition, this may reduce cloud processing time, right?
asked 26 Mar '15, 07:54
simple binarization can cause significant information loss in the image and also reduce the overall quality. This make it much harder for the OCR back end to recognise the text.
ABBYY technology uses so called "adaptive binarization" - here you can find more information: http://www.abbyy-developers.com/en:tech:insideocr:adaptive_binarisation
For a generic solution on all mobile platforms you should send (compressed) grey-scale images to the back end, so you give the algorithms more data to do intelligent binarization.
For iOS and Adrroid you might evaluate ABBYY's Mobile Imaging SDK: http://www.abbyy.com/mobile_imaging_sdk/description/
Best Regards Michael
answered 26 Mar '15, 10:52