De-Blurring a Blurred Frame Using a Sharp Frame

摘要

Sharp and blurred image frames are detected (300) and a blur kernel is estimated (302) which represents a motion-transform between their common regions. The kernel may be a point spread function describing camera motion during exposure of the blurred frame, and estimates how the frame is blurred. Using the kernel, a static region measure (e.g. static region mask) for the sharp and the blurred frame is estimated (304). A de-blurred frame is generated by replacing (306) one or more pixels of the blurred frame as indicated by the static region measure. A replacement pixel for the blurred frame may be calculated as a composite of a first pixel from the sharp frame occurring before the blurred frame and a second pixel from a second sharp frame occurring after the blurred frame. The static region measure may be generated by identifying areas of motion between the sharp and the blurred frames by comparing the blurred frame to a simulated blurred frame generated as a product of the sharp frame and the blur kernel, and may use an energy minimization function. Feature points may detect the sharp frame, e.g. using an estimation of homography.