Word Segmentation Using Center Zone Vertical Projections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing word segmentation methods struggle with accurately segmenting italic text due to vertical character overlap, which causes white gaps between words to become narrow or disappear, leading to incorrect classification of words as a single unit.
Innovation Solution
The method focuses on generating a vertical projection profile for the center zone of the text line, excluding the top and bottom zones where character overlap occurs, and uses this profile to identify and classify spacing segments into word and character spacings, thereby segmenting the text line into individual words.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vertical projection profile of the entire text line is used for word segmentation, then the method can be applied to general text, but for italic text the vertical overlap of characters causes white gaps to become narrow or disappear leading to incorrect segmentation
Solution Approach 1:
The text line is divided into three distinct zones: top zone, center zone, and bottom zone. The vertical projection profile is calculated separately for each zone, allowing the method to focus on the center zone where character overlap is minimized, thus improving segmentation accuracy for italic text while maintaining applicability to general text.
Solution Approach 2:
Different zones of the text line are assigned different functions in the segmentation process. The center zone is used for primary word segmentation due to its clarity in distinguishing word boundaries, while the top and bottom zones can be used for supplementary information or ignored when not needed, optimizing the segmentation process for each region's characteristics.
2Reliability
If connective component based method is used for italic text word segmentation, then accuracy can be improved, but the computation cost becomes relatively high
Solution Approach 1:
The method extracts only the center zone from the complete text line image, excluding the top and bottom zones where character overlap occurs in italic text. This extraction reduces the amount of data to be processed while retaining the essential information needed for accurate word segmentation, thereby improving both accuracy and computation speed.
Solution Approach 2:
Instead of processing the entire text line or using complex connective component analysis, the method applies partial action by focusing only on the center zone projection. This partial processing achieves sufficient segmentation accuracy for italic text with significantly reduced computational requirements compared to comprehensive methods.
Data Source
AI summary
A word segmentation method for segmenting a text line into word segments, which is particularly advantageous for processing italic text but can also be used for regular text. A horizontal center zone of the text line, corresponding to the vertical center parts of the characters, is used to generate a center-zone-only vertical projection profile. The center zone is determined using a horizontal projection profile, by locating the two major peaks of that profile and defining the two major peak positions as the upper and lower boundaries of the center zone. Spacing segments (white gaps) in the vertical projection profile are identified, and classified into two classes, namely character spacing (gap between characters with a word) and word spacing (gap between words). The word spacings are used to segment the text line into word segments.