What preprocessing techniques are essential in image processing pipelines for deep learning models?