Image sensors (cameras) are used in many applications, such as still cameras, video cameras, as well as in recent years, monitoring systems, automobiles, medical care, and factory automation. However, even for the latest-type digital cameras, the basic structure?optics and projective geometry?has been completely the same for film cameras invented about two centuries ago; the only difference is that the file is replaced by electric images from CCD or CMOS sensors. We propose a novel camera system that uses dynamic optical elements and new image sensing techniques utilizing a camera in this advanced image sensing project. Figure 1 illustrates the concept of the proposed camera system.
Fig.2 Advanced Camera
Fig.3 Expansion of Depth of Field
Fig.4 Depth Estimation
Fig.5 Light Field Acquisition
n this study, we propose a new imaging technique that actively controls optical characteristics for capturing an image. It is completely different from conventional cameras, since the optics of the conventional camera is usually fixed.
In advanced image sensing, we aim to solve many problems, which are difficult to eliminate with traditional images, by reviewing the camera optics and hardware in order to reconsider the best way to sensing input images for applications. Traditional cameras use passive optical elements, such as a lens and filter, which do not change the refraction index, transmittance, or polarization property. The proposed camera is realized by using special active optical devices: LCDs, LCoS, and tunable filters that change the characteristics with electrical signals. The camera can also use auto-focus and auto-iris mechanisms already part of modern cameras. The combination of these mechanisms enables us to quickly change the optical characteristics while shooting images, which allows information about scene depth, resolution, and movement to be coded and images to be multiplexed. Based on this concept, we implemented focus-sweep and programmable aperture cameras and then realized many applications, such as image deblurring, extending or controlling the depth of field, scene depth estimation, and acquisition of light-field images. We believe that this study not only proposes an image acquisition technique for particular applications, but also has the potential to greatly extend the traditional frameworks and any applications of image processing and computer vision, which can handle only images shot with normal cameras.
This work was supported by
KAKENHI Grant-in-Aid for Young Scientists (A) (21680018) and Grant-in-Aid for challenging Exploratory Research (20650023)
Microsoft Research Core 8 project