Touching letters cause incorrect word zoning and subsequently incorrect OCR.

[idrise]

When lines are close and letters on different lines touch one another, words are not bounded correctly.

Environment

• Tesseract Version: 5.0.0-alpha and 4.1 and 4.0
• Commit Number:
• Platform: ubuntu, mac osx.

Current Behavior:

level	page	block	par	line	word	left	top	width	height	conf	text
1	1	0	0	0	0	0	0	229	155	-1
2	1	1	0	0	0	59	31	137	91	-1
3	1	1	1	0	0	59	31	137	91	-1
4	1	1	1	1	0	59	31	124	36	-1
5	1	1	1	1	1	59	31	124	36	89	ankie
4	1	1	1	2	0	66	76	130	46	-1
5	1	1	1	2	1	66	76	130	46	91	ranky

Expected Behavior:

Created expected behaviour by creating a single pixel gap.

level	page	block	par	line	word	left	top	width	height	conf	text
1	1	0	0	0	0	0	0	229	155	-1
2	1	1	0	0	0	30	31	166	91	-1
3	1	1	1	0	0	30	31	166	91	-1
4	1	1	1	1	0	30	31	153	44	-1
5	1	1	1	1	1	30	31	153	44	91	yankie
4	1	1	1	2	0	33	76	163	46	-1
5	1	1	1	2	1	33	76	163	46	92	Pranky

Suggested Fix:

Not certain...

[amitdo]

The problem is in the layout analysis phase.

AFAIK, there is no solution for this issue.

[amitdo]

To solve this issue, major changes to the layout analysis module are needed.

[amitdo]

This issue is unlikely to be solved in the foreseeable future.

@stweil, I think we should close this issue as wontfix.