Text recognition is a field that has been researched and applied for many years. Text recognition
process is performed through the following main steps: The input image page will go through the
preprocessing step, then the page analysis step, the output of the page analysis will be the input
of the recognition step, and finally post-processing. The result of a recognition system depends
on two main steps: page analysis and recognition. At this point, the problem of recognition on
printed text has been resolved almost completely (ABBYY's FineReader 12.0 commercial product
can recognize printed text in various languages, recognition software of Vietnamese words in
VnDOCR 4.0 of the Hanoi Information Technology Institute can recognize with accuracy over
98%). However, in the world as well as in Vietnam, the page analysis problem remains a major
challenge for researchers. Until now, page analysis is still receiving the attention of many
researchers. Every two years in the world there is an international page analysis contest to
promote the development of page analysis algorithms. These were the motivations for the
dissertation to try researching so that they can propose effective solutions to the page analysis
problem.
In recent years, there are many page analysis algorithms have been developed, especially are
hybrid-oriented approached development algorithms. The proposed algorithms show different
strengths and weaknesses, but in general most of them still suffer from two basic errors: an error
separating a correct text area into smaller that leads to mislead or miss the information of text
lines or paragraph (over-segmentation), the aggregation error of text areas in text columns or
paragraphs together (under-segmentation). Therefore, the objective of the dissertation is to
study and develop page analysis algorithms that simultaneously reduce both types of errors:
over-segmentation, under-segmentation. The issues in page analysis are very broad so the
dissertation limits the scale of the study within the scope of text image pages written in Latin
language which particularly is English and focuses on the analysis of the text areas. The
dissertation has not proposed the problem of detecting and analyzing the structure of table
spaces, detecting image areas and analyzing logical structures. With the objectives of the
dissertation have achieved the following results:
1. Propose a solution that speeds up the algorithm for detecting background images.
2. Proposed adaptive parameterization method reduces the effect of size and font type on
the results of page analysis.
3. Proposed a new solution for the problem of detecting and using separator objects in page
analysis algorithms.
4. Proposes a new solution that separates text areas into paragraphs based on context
analysis
