Enhance handwritten text segmentation

[PlathC]

The following documentation might be useful to implement a more precise text segmentation in order to enhance the final result :

• A perceptive method for handwritten text segmentation
• Word Segmentation of Off-line Handwritten Documents
• Segmentation of Handwritten Document Images into Text Lines
• Handwritten text line segmentation using Fully Convolutional Network
• Line and word segmentation of handwritten text documents written in gurmukhi script using mid point detection technique

This one described some processing that have been used to process some pages that could be similar to our case :

• Is there an efficient algorithm for segmentation of handwritten text?

[PlathC]

• Deep Neural Networks for Large Vocabulary Handwritten Text Recognition
• Paragraph text segmentation into lines with Recurrent Neural Networks : First implementation (incomplete).
• A Steerable Directional Local Profile Technique for Extraction of Handwritten Arabic Text Lines

[PlathC]

• Is there an efficient algorithm for segmentation of handwritten text?

Tested with following result :

Python source code

[PlathC]

In order to provide uniform image to the recognition system, we may needs to deskew text line based on some known techniques :

• Measuring document image skew and orientation