I'm looking for a Java framework to help with some data mining specific to images. We have a set of historical images that I would like to categorize and classify. I'm was hoping to find something like weka http://www.cs.waikato.ac.nz/ml/weka/ or Marsyas http://marsyas.sness.net but more specific to sifting through image data to find patterns. Any suggestions?
What about using the OpenCV library for Processing? Technically, Processing is not Java, but it runs on the JVM and shouldn't be difficult to get working. OpenCV is the standard choice for computer vision, which is what you're trying to do.
I played around with Neuroph (http://neuroph.sourceforge.net) which is a neural network framework for Java, and they have a very nice image recognition tutorial that I think everybody should read, but specially if you're dealing with image recognition: http://neuroph.sourceforge.net/image_recognition.html
Have a look at OpenCV with JavaCV libraries from Google. Simply use function from that library to find whatever patterns you want.
It'll be a great surprise for me if you can find such library.
I can advice you to find some library with ART neural network implementation. Then you can add some photos to each category you want to recognize and ART algorithm will try to recognize all other photos.
来源：https://stackoverflow.com/questions/2062618/java-framework-for-image-pattern-recognition