Squish Options for Tesseract

Last edited on

Documentation for Tesseract 4

Tesseract is an open source project. The documentation for Tesseract 4 can be found here:

Tesseract 4 Documentation

Squish Options for Tesseract

Squish currently offers a few options for configuring OCR performed by Tesseract. Most of these options are Tesseract specific and are directly based on functionality provided by Tesseract itself:

Image Processing

These are pre-processing options.

These are used to convert the image into black & white pixels before OCR happens.

Note that None is not the same as Default.

Flavor

A property of individual language files shipping with Tesseract.

Language

Language specifies the expected language of the text that should be recognized by Tesseract.

Mode

This parameter indicates which LTSM model (also known as "tessdata") is used to do OCR, depending on what the priority is: speed/performance or accuracy.

The choice, in theory, is between tessdata_fast, tessdata_best or just tessdata.

The Tesseract installer provided by the Qt Group currently only supports "tessdata_fast", which is denoted by the value "(Default)".

Segmentation

This relates to Tesseract's "Page segmentation method", which is documented here:

Page segmentation method

Postprocessing

The post-processing option determines whether the built-in algorithm provided by Squish should be used to make assumptions about the layout and structure of the text that Tesseract has found, which relates to arranging lines that belong together into logical groups to enable multi-line search.

When set to "default", the built-in algorithm provided by Squish is enabled. When set to "none", the text found by Tesseract is not being post-processed.