The present invention relates to a 3D image capturing structure, comprising a lens body. A lens group, a prism group and a photosensitive element are arranged in the lens body. The prism group comprises a mounting seat, the inner circumference of the mounting seat is connected to a sliding seat by means of an angle adjusting mechanism, and the sliding seat may rotate under the action of the angle adjusting mechanism so that an included angle is formed between the axis of the sliding seat and the axis of the mounting seat. The inner circumference of the sliding seat is connected to a mounting plate by means of a moving mechanism, and the mounting plate may move along the axis direction of the sliding seat under the action of the moving mechanism. One or more polygon prisms are mounted on the mounting plate and used for collecting 3D images or light field images. Each polygon prism comprises at least four edges. Two parallel outer surfaces of each polygon prism form an incident plane and an emergent plane respectively, and the other outer surfaces of the polygon prism are coated with shading coatings. According to the present invention, one polygon prism only forms one image, so that the total amount of emergent light and the angle of the light may be conveniently adjusted.