As is known as Depth from Defocus, we can compute a 3D range image from multi-focus images, a set of defocused images taken with different focus values (Fig. 9)[5].
In [6], we proposed algorithms to detect edges and measure 3D depth from multi-focus images. In this project, we improved the algorithms and developed a multi-focus video camera which can capture three defocused images simultaneously. Fig. 10 shows the imaging device of the camera, where three CCDs are located at different positions along the optical axis. Experiments using this camera showed that while the accuracy of depth sensing varies depending on camera parameters, 0.3% relative depth errors can be attained under optimal environments[7]. This 3D range sensing method is passive, real time, single camera with no correspondence problem, and moreover measurable of very wide range of depth by changing camera parameters (focal length, focus value, and iris size). Just before this workshop, we devised a still new method of depth measurement by the multi-focus camera. [8] describes its technical details.
Figure 9: Multi-focus images. In this camera model,
the position of the image plane is shifted according to the focus value.
Figure 10: Multi-focus camera. An ordinary 3CCD video camera is reformed
to the multi-focus camera.
