Multi-View Background Subtraction for Object Detection

Consider a popular tourist destination shown in Figure 1. How can we exploit the large set of photographs available online depicting this same general location in order to better understand the content of this particular image? It is useful to divide scene components into two categories: dynamic objects such as people, bikes, cars, pigeons or street vendors that move about and are likely to only appear in an image taken at a particular time, and static backgrounds such as buildings, streets, landscaping, or benches that are visible in many different images taken in the same location.