Crash effects of table lines on recognition accuracy

[sxgkwei]

Environment

• Tesseract Version: 4.1.1-rc2 , language data: chi_sim.traineddata
• Commit Number:
• Platform: Windows 7 64-bit

Current Behavior:

When there is a table on the picture, it cannot be recognized. The content in the table is basically completely confusing.

Expected Behavior:

The text in the form can be recognized normally

Suggested Fix:

Able to fix the misjudgment of horizontal / vertical / oblique table lines when recognizing converted text

[amitdo]

The issues which might make it hard for Tesseract to recognize this image:

1. A stamp which probably confuses the table detection.
2. Mixing two writing directions in one block: left to right and top to bottom (Tesseract considers a table as one block).