Predicted Classification Response to Label Matrix

Hello everybody

I have a question that seems pretty basic but I am not sure if there is an answer online for it. Is it possible somehow to get a label matrix back from the prediction response of a classifier (decision tree in my case). I have the values (and the string names) given from the classifier and I would like to assign them to a label image. I have already created a label image before (with bwboundaries()) for a previous step.

So the overall question is how is possible to connect the prediction of a classifier with a region labeled image ?

Thank you in advance

[Dimitris M]

Please if somebody has any suggestion please give me a hint. I am not looking for coding solutions just a thinking workpath over this problem !

[Tom Lane]

I would need more information to understand the problem. If your training data consist of rgb values and labels, and you have a new image with rgb values, then I would expect the classifier to produce the same values. Please explain what you have, what you are doing, and what you need.

[Dimitris M]

OK. So I have a label images produced with the function bwboundaries(). I have used this labeled image region (and its true color image) to classify it according to a training set I created.

The problem is that the values in the labeled image are not the same as the numbers the classifier assigns to the classification result. My question is how from the classification numbers assigned by the classifier I can go back to a label image (so I can color it according to the result). Is there a way to move backwards after classification and visualize the image objects according to their class ?

If my question is still not clear to you let me know, so I can explain further.

Thank you in advance

[Tom Lane]

Sorry, I really don't know much about image processing. I don't understand what data you would use to create the classifier and what output you would hope to get in doing the prediction. Maybe someone else with more image processing knowledge can help.