tesseract-ocr/tesseract

Small extra blocks with a single letter gets split off of bigger text blocks

Open

#2634 opened on Aug 30, 2019

View on GitHub
 (1 comment) (0 reactions) (0 assignees)C++ (74,090 stars) (10,622 forks)batch import
help wantedlayout analysis

Description

Environment

  • Tesseract Version: tesseract 5.0.0-alpha-357-gdc907 leptonica-1.76.0 libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 1.5.2) : libpng 1.6.36 : libtiff 4.0.10 : zlib 1.2.11 : libwebp 0.6.1 : libopenjp2 2.3.0 Found AVX2 Found AVX Found FMA Found SSE
  • Commit Number: dc90741f1b8f37e8d1a0c919bb679f455bd39633
  • Platform: Linux jk-XPS-13 5.0.0-25-generic #26-Ubuntu SMP Thu Aug 1 12:04:58 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux

Current Behavior:

This page of a book gets recognized perfectly except for „die Sache selbst“ (end of third line), which becomes „die Sache selbst‘ (single quote). The other single quote becomes another block containing only the very small character "C" testcase

I'm sorry I could not provide a cropped test image, but for smaller regions the problem disappears.

I'm calling tesseract with default parameters: tesseract testcase.png - -l deu

When called with psm 6 (single uniform block of text) it works, but I don't want to loose the layout information. tesseract testcase.png - -l deu -psm 6

This is of course a minor bug, but maybe it's also easy to fix. It happens like one time in hundred pages. Sometimes footnote numbers get lost the same way. The problem appears at least with tesseract 4.0 / 4.1 / master and in all oem modes.

Expected Behavior:

tesseract should not split of single chars in extra regions

Suggested Fix:

Maybe padding the recognized blocks a bit?

Contributor guide

Tech stack
cpp
Domain
backend
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-2 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
C++ proficiencyTesseract build setupLayout analysis concepts
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
35
Research direction
Examine the block segmentation code in Tesseract, focusing on how text blocks are split. The main files to review are 'ccmain/paragraphs.cpp' and 'ccstruct/blobs.cpp'. The issue suggests that small blocks (single characters) are incorrectly separated; consider adding a merging step or increasing padding. No maintainer response yet; consider asking for a reproducible image or creating one.