In scene analysis, the availability of an initial background model that describes the scene without foreground objects is the prerequisite, or at least can be of help, for many applications, including: video surveillance, where an accurate estimate of the background is essential for high quality moving object detection based on background subtraction; video segmentation, where the background provides rich information to extract foreground objects; video compression (or video coding), where the estimated background represents redundant information that can be suppressed; video inpainting (or video completion), whose techniques try to fill-in user defined spatio-temporal holes in a video sequence using information extracted in the existent spatio-temporal volume, according to consistency criteria evaluated both in time and space; privacy protection for videos, whose main aim is to avoid the infringement on the privacy right of people taken in the many videos uploaded to video sharing services, such as YouTube, that may contain privacy sensitive information of the people; computational photography, where the user wants to obtain a clean background plate from a set of input images containing cluttering foreground objects. The aim of SBMI2015 (sbmi2015.na.icar.cnr.it/) is to bring together researchers interested in scene background modeling and initialization (also known as bootstrapping, background estimation, background reconstruction, initial background extraction, or background generation) in different application areas, in order to: disseminate their most recent research results, advocate and promote the research into scene background modeling and initialization, discuss rigorously and systematically potential solutions and challenges, promote new collaborations among researchers working in different application areas, share innovative ideas and solutions for exploiting the potential synergies emerging from the integration of different application domains.