I'm using the Java OCR library and have been having a difficult time getting accurate results out of small PNG images with a relative resolution of 300 dpi. I've attached an example image to this entry. In my latest iteration I've configured my OcrEngine with three filters; MedianFilter(3);, GaussBlurFilter();, and RemoveNoiseFilter();. I've also run tests excluding and including various combinations of the three, including different values for the MedianFilter. But this latest iteration seems to provide the best results, which are still not great. Here's the output from the attached file;
iFromUnderraduateClasses Under raduate Classes
Chemistry: familiar with titrationscvolumetric:colorimetric. gravimetric
analysisismall-scale synthesis: d istillation ccompou nd isolation .
identification. purification}extraction: chromatography (gas. paper. thin
layer) cd i I ution s crefractive i ndex
Biology: familiar with microscopy:tissue culturingiblood typing: dissectionc
protqin. DNA. RNA isolationcbacteriologic analysis: acid-fast, gram stainsi
immu noprecipitation cwestern blots
Any assistance in getting this to produce consistent and accurate results would be greatly appreciated.
Thank you for your inquiry and sharing sample.
This is to update you that we have investigated the issue at our end. Initial investigation shows that the issue persists. The issue has been logged into our system with ID OCRJAVA-696. Our product team will further look into it. We will update you via this forum thread once there is any update is available on this issue.