Re: [orca-list] document recognizing

["Mgr. Janusz Chmiel" <janusz chmiel volny cz>]

Because sighted users can use graphical capabilities of OCR software to 
apply some effects to improve The quality of text output. Issue is, that 
Tesseract, which support many languages have not been programmed to 
support high or very low resolution. 300 DPI is The bestvalue. And is 
some user uses camera, which have much more lover or higher resolution, 
output is not satisfied. Next problem is, that user, who do not see at 
all can not take picture of A4 paper without guidance. So app would 
hadto give him instructions in realtime. Voice isslow, so only sound 
tones could enable this.

Unfortunately OCR software is based on complex mathematical analysis, 
memory manipulations andallocations and this is The reason, why this 
kind of software if it must create good quality will be never free or 
cheap to buy.
I think, that The way is to discuss with LIOS developers, if there are 
some Python or .so libraries with Python programmers library, which can 
analyse graphical input and if it is programmatic ally possible. 
Algorithm could analyse The input resolution. I do not know, if 
mathematical algorithms, which are able to change resolution of input 
graphical files do not have negative effect on The image quality. And I 
Am afraid,that for example, 150 DPI image can not be safely transformed 
to 300 DPI output file.