TextFusion: Unveiling the Power of Textual Semantics for Controllable Image Fusion