Segmentation and validation of commercial documents logical structure

Abstract

The main objective of the work is to present an approach to extract and validate the logical structure from the images that compose a commercial document. The nearest neighbor rule algorithm was used for labeling the elements, and the Run Length Smoothing Algorithm (RLSA) was used to segment the image of a commercial document of the type letter, official letter or memo. The most common classes considered are: date, logotype, text body, signature, addressee, invocation and greeting. The labeling of the elements is accomplished using the nearest neighbor rule algorithm with a vector comprising 28 characteristics. The accomplished study presented a good result for the classification of elements on commercial documents. It was created and used a base composed of 283 images of commercial documents in 256 gray levels for document element classification.