Skip to content

Optical Character Recognition in Chinese Letter Manuscripts (C++)

Notifications You must be signed in to change notification settings

sebastianstarke/OCR

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

17 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Optical Character Recognition in Chinese Manuscripts

Description

This project implements OCR (Optical Character Recognition) for chinese manuscripts. OCR in general is a popular research problem in image processing and pattern recognition. In particular, chinese letters make it even more complex by their high amount of variation in rotation, scale, line strength and several other features. The goal of this project was to determine whether two manuscripts were written by the same author. Solving this problem, the algorithm first performs several morphological transformation which then are folllowed by aligned intensity projections in order to obtain a reliable segmentation of letters. Afterwards, 10-dimensional feature vectors for all detected letters are computed and classified regarding their similarity by a slightly modified version of the k-Nearest-Neighbor algorithm modeling a dual-likelihood classification. Ultimately, the majority vote of detected matching letters among all gives rise to the confidence rate of two manuscripts being written by the same author. The project was implemented using Visual Studio with C++ and the OpenCV framework. It was part of the Master's course "Image Processing" at the University of Hamburg and has been awarded as the most robust, adaptive and accurate approach.

The pictures below illustrate the segmentation and detection process, followed by a recognition for writer-similarity identification.

Demo

#1 - Start WINDOWS (or anything that can execute .exe)

#2 - Open cmd and navigate to the OCR.exe folder

#3 - Type: OCR.exe

#4 - imagepath1 & imagepath2 -> paths to your .jpg manuscript

#5 - k -> parameter for kNN classification

#6 - sigma -> parameter for kNN decision-confidence threshold

#7 - output -> 0 or 1 0 if you only want to see the results 1 if you also want to see the processing steps and probability distributions

#8 - Example: OCR.exe front/3.jpg back/3.jpg 50 0.25 1

About

Optical Character Recognition in Chinese Letter Manuscripts (C++)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages