A system to find text areas according to textural features in digital document images

Written and printed documents that are transferred to digital platform, provided an easy way for gathering, storing, updating and sharing of information but increased research in page layout analysis and text area extraction. In this work, a system that finds text areas in digital document images, was developed with evaluation of the textural feature difference between image region and text region that is composed of text characters. In document images high frequency and low frequency components were extracted by applying single value proportional Gabor filter that is created due to angle values which text characters are sensitive to. With the idea of high frequency components are probable text characters, some high frequency components such as lines, windows etc. were eliminated by connected-component analysis. Different from existing methods, an elimination named character tracing was done on probable text characters with evaluation of text character?s continuity and neighborhood. Document model were created for optical character recognition process after finding text characters with character tracing method. In the designed system which was developed due to suggested approach, the process of text area extraction was speed up by using of one value rationed Gabor filter in Gabor filtering step. Character tracing method which is a new approach, increased the success rate of system in process of text characters determination. With this system, successful results are obtained in documents which have complex page layouts and text areas which are composed of different alphabet?s characters such as Latin and Cyrillic. Keywords: Text area extraction, Gabor filtering, Multi-channel filtering, Page layout analysis, Character tracing.