Numerical Methods for Blind Image De-Convolution

摘要

Despite all the advances in digital photography, motion blurring is still one of the most common causes of blurred pictures caused by camera movement during exposure time. A motion-blurred picture is usually modeled by f=p * g+n, where * denotes the discrete convolution operator, f denotes the blurry picture, g denotes the sharp one, n denotes image noise, and p denotes the blur kernel determined by the relative motion between the camera and the scene. Modern cameras address this issue by using image stabilization, which controls mechanical actuators that shift the sensor or lens element during the exposure to compensate for motion of the camera and thereby reduce mage degradation. Unfortunately, image stabilization only addresses very modest camera shake.