Imagine a beautiful picture, perhaps a sunset or a portrait, destroyed by patronising words on top of it. If a picture requires words to render it comprehensible the image is poorly chosen. If the intent of the image is clear, adding text is merely patronising, not explanatory. Images are complete in themselves.
The rationale behind annotation of images in presentations is unclear. Templates infer the need for a title on every slide. The misunderstood belief that text is required to make sense of the spoken work often persists once dependence on bulletpoints has passed. Signposting is entirely superfluous: as valuable as signposting a signpost.
Text can add or change meaning in an image. That effectively combines the two elements as a whole. If the import of this whole would remain unaltered by removal of one component, that part is superfluous. Annotation of complex images adds only further complexity. If an image is so complex as to require explanation that can only be verbal.
An image should have enough strength to stand unaided. If a title is required, the image is weak and unhelpful. Duplicating the message of an image is patronising.