Home

Java OCR is a suite of pure java libraries for image processing and character recognition. Small memory footprint and lack of external dependencies makes it suitable for android development. Provides modular structure for easier deployment.

actual and olny release exists as bundle of jars in maven central repository, and you can just reference it in your project. We are working to provide downloadable bundle

How it works

JavaOCR is pure java suite of libraries for image processing and recognition. It uses (for now) invariant moments and cluster analysis for character recognition and pattern matching. Deatiled explanations of this techinque are provided here [image recognition using moments]

Project structure

JavaOCR is split into smaller components files so you can pick up what you really need

• core - contains basic image definitions, core filters (like sauvola binarisation) , traversal classes and slicing code, integral image creation. suppors ineteger byte and float images
• plugins - contains various plugin functionality
  • awt - integration with awt
  • cluster - cluster analysis classes, including euclidian and mahalanobis metrics
  • fir - finite impuls response filters like median filter ( obsolete, better implemented via integral images )
  • moment - filters to compute invariant moments over images including Hu-Moments. Hu-Moments are well described but not very useful for character recognition as thez provide rotational invariance and will likely mix up 2 and 5. You will have to choose your own moment set
  • morphology - mophology operations (erosion / dilation / opening / closing)

Compiling

To compile project from source, get sources and issue mvn clean install frmo top level directory. This will install all project jars into local maven repository. Released version is also available from sonatype nexus repository and shal be synched to maven central.

Using JavaOCR in your projects

JavaOCR is pure java library and can be used everywhere java is running. demos/ folder contains android showcase apps. Read more about compiling and running them: [android demos]

Project admins

Project Admins:

• Ronald B. Cemer
• Konstantin Pribluda
• William Whitney

[Ronald B. Cemer]

This project is currently abandoned, and will not be maintained any longer. Comments on this wiki will also no longer be read or approved. If you want to take the source code elsewhere, for example, GitHub, and work on this project further, feel free. However, this is very OLD technology. Modern OCR uses AI, and is far more flexible and effective.

Additionally, Java is no longer the cutting-edge language it was 20 years ago. New applications should be written in Go, Rust, or another modern language which compiles to native code. As a result, this project is not something in which the original author (Ronald B. Cemer) would recommend investing in further work.