PENGENALAN TULISAN TANGAN DENGAN PERBAIKAN GORESAN MENGGUNAKAN INTERPOLASI BEZIER DAN SMOOTHING

Abstract

This system can process images in the form of letters handwriting into digital text writing and at the time of processing, the damage from the handwriting image repaired first in order to be a form letter similar to the actual shape of the letter so that it can add recognition accuracy in using Bezier interpolation and smoothing scratches on image which is then carried handwriting feature extraction process with Global Histogram and then calculated the distance of the feature extraction with Manhattan distance or Euclidean Distance.

Test results using Manhattan Distance that can achieve 80% recognition rate on number reognition, greater than the Euclidean Distance that can achieve 70% recognition rate on number recognition. The successful recognition of many affected by the similarities in the character of handwriting. In particular the test database and test images are split between uppercase, lowercase and numbers, it appears that the percentage of the recognition  improved quite a lot.