What does ImageDataGenerators argument featurewise_center do?

Hey guys, while doing a project in Deep Learning I stumbled across a documentation in tensorflows ImageDataGenerator: tf.keras.preprocessing.image.ImageDataGenerator

under the argument featurewise_center it is mentioned it will ‘Set input mean to 0 over the dataset, feature-wise.’ I do not fully understand the functionality here. What mean is set to 0? As far as I understood using ImageDataGenerator I can create an iterator which I then cann pass to my neural network. What mean is set here? I thought the data gets passed in as array of pixel values, there should be no meaningful mean to generate here?

Thanks in advance for your answer!