by Stommel, Martin and Frieder, Gideon
Abstract:
In previous work we introduced a function that estimates the legibility of a binary text image. The function is trained in a supervised way to distinguish between different qualities of binarisation results. A framework is introduced that uses this function to automatically choose filter parameters to enhance the legibility of historic documents without user interaction. Here, we present the results for 93 historic documents. For 80 per cent of the documents appropriate binarisation thresholds could be achieved. The results encourage the application of the method to more sophisticated filters.
Reference:
Stommel, Martin and Frieder, Gideon, "Evaluation of the Legibility Estimation for 93 Historic Documents", Technical report, TZI Universität Bremen, no. 57, 2011.
Bibtex Entry:
@TECHREPORT{Stommel2011,
author = {Stommel, Martin and Frieder, Gideon},
title = {Evaluation of the Legibility Estimation for 93 Historic Documents},
institution = {TZI Universit{\"a}t Bremen},
year = {2011},
type = {TZI Technical Report},
number = {57},
month = {February17},
abstract = {In previous work we introduced a function that estimates the legibility
of a binary text image. The function is trained in a supervised way
to distinguish between different qualities of binarisation results.
A framework is introduced that uses this function to automatically
choose filter parameters to enhance the legibility of historic documents
without user interaction. Here, we present the results for 93 historic
documents. For 80 per cent of the documents appropriate binarisation
thresholds could be achieved. The results encourage the application
of the method to more sophisticated filters.},
owner = {pmania},
timestamp = {2012.11.06}
}