User-guided pedestrian and object removal

Street-level image collections now exist for a number of locations around the world. While the focus is on capturing streets and buildings, pedestrians are commonly captured. Pedestrian faces are identified and blurred in different system to address privacy concerns but leave their possibly identifiable bodies intact. In this paper, we present a user-guided approach to fully remove pedestrians and undesirable scene objects. Both are fully removed given a user created removal mask with our system's gui and using successive image frames with an image inpainting-inspired approach.