Abstract: In multi-volume visualization, adjacent views have difficulties in establishing spatial correspondences and recognizing differences between features, while integrated views may cause occlusions between features and make the visualization not easy to understand. This paper presents a semantic lens, a tool for visualizing and inspecting multiple volumes interactively. First, a semantic list is designed using multiple attributes of volumes to establish the relationship between the semantic descriptions, such as size and shape, and multi-dimensional transfer functions, and then the user can interactively define the features inside and outside the lens on demand. To connect features from outside to inner, the correct perception of the fused color is evaluated based on the color perception theory. Then, the structure fusion or color fusion is recommended to integrate different features in the transition region. Finally, the directions of rays only behind the lens are modified and go through the focal point. Thus, the feature at a specific depth will be magnified while other features before the lens remain the original size. We use the semantic lens to interactively inspect and analyze three different types of data: multi-feature single volumes, multi-modal and multi-variant volumes. A large number of comparison results further demonstrate the effectiveness and practicability of the tool, along with the evaluation and feedback from users.