Single-Shot Image Deblurring with Modified Camera Optics

The recent rapid popularization of digital cameras allows people to capture a large number of digital photographs easily, and this situation makes automatic avoidance and correction of “failure” photographs important. While exposure and color issues have been mostly resolved by the improvement in automatic corrective functions of cameras, defocus, motion, and camera shake blur can be handled only in a limited fashion by current cameras. Camera shake blur can be alleviated by an anti-camera shake mechanism installed in most cameras; but for focus, although a particular scene depth can be focused with an auto-focus function, objects at different depths cannot be captured sharply at the same time. Moreover, defocused images can often result due to the failure of auto-focusing. In addition, blur caused by object motion, i.e., motion blur, is only avoided by increasing the shutter speed and sensor sensitivity when a camera detects motion in a scene. This dissertation proposes a method for removing defocus and motion blur for digital cameras. Since deblurring is generally an ill-posed problem, and hence an image processing approach alone has limitations, the proposed method includes modifications of camera optics. In this regard, this dissertation pursues low cost and compact implementation, aiming at applications to consumer products. That is, small modifications to existing cameras or mechanisms that can be directly derived from existing ones will be adopted. In order to set a baseline performance achievable without modifying camera optics, this dissertation first proposes an image deblurring method that is purely based on an image processing approach, which consists of fast image deconvolution for efficient deblurring, and local blur estimation for handling spatially-varying blur. Additionally, a set of intuitive user interfaces are provided with which the user can interactively change the focus settings of photographs after they are captured, so that she/he can not only obtain an all-in-focus image but also create images focused to different depths. For removal of defocus blur, a method is proposed for estimating the defocus blur size in each image region by placing red, green, and blue color filters in a camera lens aperture. As captured image will have depth-dependent color misalignment, the scene depth can be estimated by solving a stereo correspondence problem between images recorded with different wavelengths. Since the depth is directly related to the defocus blur size, deblurred images can be produced by deconvolving each region with the estimated blur size. The modification requires only inexpensive color filters. For motion blur removal, this dissertation proposes to move the camera image sensor circularly about the optical axis during exposure, so that the attenuation of high frequency image content due to motion blur can be prevented, facilitating deconvolution.