recognizes more characters than present

[abieler]

• Tesseract Version: 4.00 (with tessdata_best, lstm-only, lang=deu+eng)
• Platform: arch linux 64 bit

Current Behavior:

With psm=7 (single text line):
For the thresholded image below recognizes "2427. 50" instead of "24. 50", for the original image recognizes "242. 50".

For psm=1 or 3 it complains about "empty page" for both images.

[Shreeshrii]

@zdenop Label with

4.0x
Accuracy

[Shreeshrii]

With current code, correct output with tessdata and tessdata_best with psm 6 or 7 or 8 for first image.

tesseract 4.1.0-rc1-255-g332a1
 leptonica-1.76.0
  libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 1.4.2) : libpng 1.2.54 : libtiff 4.0.6 : zlib 1.2.8 : libwebp 0.4.4 : libopenjp2 2.3.0

*****  1362-1.jpg OEM 1 LANG eng TESSDATA tessdata_best
**** PSM 3 ****
Empty page!!
Empty page!!
**** PSM 6 ****
42.50
**** PSM 7 ****
42.50
**** PSM 8 ****
42.50

*****  1362-1.jpg OEM 1 LANG eng TESSDATA tessdata
**** PSM 3 ****
Empty page!!
Empty page!!
**** PSM 6 ****
42.50
**** PSM 7 ****
42.50
**** PSM 8 ****
42.50

*****  1362-1.jpg OEM 1 LANG eng TESSDATA tessdata_fast
**** PSM 3 ****
Empty page!!
Empty page!!
**** PSM 6 ****
A>. 50
**** PSM 7 ****
A>. 50
**** PSM 8 ****
a>. 50

*****  1362-2.jpg OEM 1 LANG eng TESSDATA tessdata_best
**** PSM 3 ****
Empty page!!
Empty page!!
**** PSM 6 ****
A422. 50)
**** PSM 7 ****
A422. 50)
**** PSM 8 ****
42. 50)

*****  1362-2.jpg OEM 1 LANG eng TESSDATA tessdata
**** PSM 3 ****
Empty page!!
Empty page!!
**** PSM 6 ****
422. 50)
**** PSM 7 ****
422. 50)
**** PSM 8 ****
42. 550)

*****  1362-2.jpg OEM 1 LANG eng TESSDATA tessdata_fast
**** PSM 3 ****
Empty page!!
Empty page!!
**** PSM 6 ****
A>. 50
**** PSM 7 ****
A>. 50
**** PSM 8 ****
427.50

[daviddphillips]

Here is an example of double characters when english language is used, and correct when another language (french) is used. I've tried the fast and best tessdata and also different versions of tesseract. I've seen this behavior with lots of images, specifically when there is a mixture of letters and numbers in a word.

The file is called EVO_Double.png:

tesseract EVO_Double.png stdout -l fra
Violation # N/A-56890 | License Plate EVO11B — Amount Due 78.75 Payment Due By 05/08/2019

tesseract EVO_Double.png stdout -l eng
Violation # N/A-56890 License Plate EV0O11B Amount Due 78.75 Payment Due By 05/08/2019

[stweil]

Latest Tesseract and release 4.1.1 recognize Violation # N/A-56890 License PlateEV011B Amount Due 78.75. Payment Due By 05/08/2019.

[woodjohndavid]

I have just created pull request #4211 which I consider to be an improved solution for diplopia.

I encourage everyone on this trail to try this out and test it with as broad a range of cases as possible.

Note by the way, there are some new configuration values that can only be set in code as things stand. These configuration values are:

bool kRemoveDiplopia - if true, enables diplopia removal functionality. If false, my changes have no effect
int kMaxDiplopiaGap - maximum number of timesteps apart to be considered diplopia, default 2

Obviously if my diplopia change is of value, then these configuration items should be made into settings.