How to train the classifier (using features extracted from images)?

Use the fitctree fucntion to create a classification tree based on the training data:

tModel = fitctree(xTrain, yTrain);

See what you can do with tModel by looking at its methods:

methods(tModel)

The resulting tree can be visualized with the view() function:

view(tModel, 'mode', 'graph');

New observations can be classified using the predict() function:

yPredicted = predict(tModel, newX);

The TreeBagger() function uses bootstrap aggregation ("bagging") to create an ensemble of classification trees.

tModel = TreeBagger(50, xTrain, yTrain); % Create new model based on 50 trees.

This is a more robust model.