INTER-LINE DISTANCE ESTIMATION AND TEXT LINE EXTRACTION FOR UNCONSTRAINED ONLINE HANDWRITING
Eugene H. Ratzlaff · 2004
Introduction One of the first steps often required in applying machine recognition to unconstrained handwriting is the identification and extraction of each line of text from among other lines. The purpose of text line extraction is to prepare data to meet the requirements of succeeding processing steps such as size normalization, word segmentation, or feature extraction. These steps typically require the data to be no more than a single row of characters. The goal of text line extraction is to assign correctly each stroke or component to its appropriate text line so that each isolated line may be passed in turn to the following analysis stage. The task is made difficult by the fact that data frequently contain undulations and shifts in the baseline, baseline skew, baseline-skew variability, character-size variability, sparse data, skipped lines and inter-line distance variability. Several of these issues have been addressed in much of the recent work focused on text line extract