ImageDataGenerator option conflict in test suite

[rossbar]

There are several instances in the test suite which create keras_preprocessing.image.ImageDataGenerator instances where both zca_whitening and featurewise_std_normalization are set to True, for example:

deepcell-tf/deepcell/image_generators/image_generators_test.py

Lines 484 to 490 in 17b54eb

	generator = image_generators.SampleDataGenerator(
	featurewise_center=True,
	samplewise_center=True,
	featurewise_std_normalization=True,
	samplewise_std_normalization=True,
	zca_whitening=True,
	data_format='channels_last')

However, when zca_whitening is set, featurewise_std_normalization is ignored.

Is there a preference for one config over the other? For example, if zca_whitening is preferred, we could simply set featurewise_std_normalization to False or vice-versa. Alternatively if both states are important, we could parametrize on these configs.