Object-layout-aware image retrieval for personal album management

Abstract

This demo shows a real-time object-layout-aware image retrieval system for personal album management. The query of the system is image's object layout and the system retrieves images based on the layout similarity of concerned objects in the query. Using automatic object localization algorithms, we propose to index the personal album with the object positions and scales in the 2D image plane, which can reflect their layout in 3D scene and hence be used to measure the object layout similarity. To query from the database, the users need describe their query by placing objects onto a canvas and configuring the position and scale of the objects. Then the system searches the album database and returns photos with the similar object layout. This system can facilitate people in finding the photos of which they have impression on the layout of certain objects. © 2012 Springer-Verlag.