Fusion for Image Restoration

Filip Sroubek
UTIA - Institute of Information Theory and Automation, Czech Repubic

Due to imperfections of imaging devices (optical degradations, limited resolution of CCD sensors) and instability of observed scenes (object motion, media turbulence), acquired images are often blurred, noisy and may exhibit insufficient spatial and/or temporal resolution. If multiple images of the scene are available, a reliable estimation of the original image can be achieved by image fusion. The talk will present one particular type of fusion, namely fusion for image restoration, which consists of blind deconvolution and resolution enhancement. Multiframe deconvolution estimates and removes blurring, and the spatial resolution of images is increased by so-called superresolution. We have propose a unifying system that simultaneously estimates blurs and recovers the original undistorted image, all in high resolution, without any prior knowledge of the blurs and original image. We accomplish this by formulating the problem as constrained least squares energy minimization with appropriate regularization terms, which guarantees a close-to-perfect solution. The performance of the proposed superresolution method will be demonstrated on several examples, such as car license plate recognition and face recognition. A live demo showing fusion with a standard webcam connected to a laptop is planned for the end of the talk.