image.textlinedetector

The image.textlinedetector R package detects text lines in digital images and segments these into words.

Objective of the package is to more easily plug the text lines in Handwritten Text Recognition modelling frameworks like the one explained in this document

The algorithms in this R package implement the following techniques:

  1. An Implementation of a Novel A* Path Planning Algorithm for Line Segmentation of Handwritten Documents paper link
  2. A Statistical approach to line segmentation in handwritten documents paper link
  3. A new normalization technique for cursive handwritten words paper link

More descriptions of technique 2 can be found in this document

Installation

install.packages("opencv")
install.packages("magick")
install.packages("image.binarization")
remotes::install_github("DIGI-VUB/image.textlinedetector")

Look to the documentation of the functions

help(package = "image.textlinedetector")

Example

Based on the paper An Implementation of a Novel A* Path Planning Algorithm for Line Segmentation of Handwritten Documents

library(opencv)
library(magick)
library(image.binarization)
library(image.textlinedetector)
#path <- "C:/Users/Jan/Desktop/OCR-HTR/RABrugge_TBO119_693_088.jpg"
path  <- system.file(package = "image.textlinedetector", "extdata", "example.png")
img   <- image_read(path)
img   <- image_binarization(img, type = "su")
areas <- image_textlines_astar(img, morph = TRUE, step = 2, mfactor = 5)
areas <- lines(areas, img, channels = "bgr")
areas$n
areas$overview
areas$lines
areas$textlines[[2]]
areas$textlines[[4]]
combined <- lapply(areas$textlines, FUN = function(x) image_read(ocv_bitmap(x)))
combined <- do.call(c, combined)
combined

Based on the paper A Statistical approach to line segmentation in handwritten documents

library(opencv)
library(magick)
library(image.binarization)
library(image.textlinedetector)
path   <- system.file(package = "image.textlinedetector", "extdata", "example.png")
img    <- image_read(path)
img
img_bw <- image_binarization(img, type = "isauvola")
areas  <- image_textlines_flor(img, light = TRUE, type = "sauvola")
areas$overview
areas$textlines[[6]]
areas  <- lines(areas, img_bw, channels = "gray")
textwords <- image_wordsegmentation(areas$textlines[[10]])
textwords$n
textwords$overview
textwords$words[[2]]
textwords$words[[3]]
combined <- lapply(textwords$words, FUN = function(x) image_read(ocv_bitmap(x)))
combined <- do.call(c, combined)
combined

DIGI

By DIGI: Brussels Platform for Digital Humanities: https://digi.research.vub.be