Compare text recognition for different settings

[winkelement]

Testfile: Scanned PDF, 300 dpi, Calibri 11pt

(jpeg Preview)

[winkelement]

Using PHP similar_text to calculate percentage of similarity between original text and OCR result.

[winkelement]

Convert options (png): -colorspace gray -type grayscale -density 300 -geometry 1024 -crop 0x0+0+0 -quality 90 -trim -deskew 40% -normalize -adaptive-sharpen 0x1
tesseract options: -l deu -psm 3
Match: 95.69 %

[winkelement]

Convert options (jpg): -colorspace gray -type grayscale -density 300 -geometry 1024 -crop 0x0+0+0 -quality 90 -trim -deskew 40% -normalize -adaptive-sharpen 0x1
tesseract options: -l deu -psm 3
Match: 95.98 %

[winkelement]

Convert options (tif): -colorspace gray -type grayscale -density 300 -geometry 1024 -crop 0x0+0+0 -quality 90 -trim -deskew 40% -normalize -adaptive-sharpen 0x1 -depth 8
tesseract options: -l deu -psm 3
Match: 95.83 %

[winkelement]

Convert options (jpg): -colorspace gray -type grayscale -density 300 -geometry 1024 -crop 834x535+97+93 -quality 90 -trim -deskew 40% -normalize -sharpen 0x1
tesseract options: -l deu -psm 3
Match: 96.81 %