APPARATUS AND METHOD FOR DIVIDING DOCUMENT IMAGE BASED-ON CLUSTERING

摘要

PURPOSE: An apparatus and method for dividing document image based-on clustering is provided so that regions of letters can divide and extract regardless of a size, a form, and the gradient of rows of the letter from letters having no a special structure by dividing the inputted document image into letter regions. CONSTITUTION: An apparatus and method for dividing document image based-on clustering includes an input device (10) for inputting document image. The input device (10) scans document by a scanner and converts as a digital image. An object representation device in an image (20) represents the object of the inputted image as a proper form which is suitable to clustering. A clustering initialization device (30) initializes clustering by using the representation result, i.e., produces a distance matrix and an initial cluster. A clustering device (40) executes clustering by the distance matrix, starting from the initial cluster.