Co-sponsored by: The Department of Electrical and Computer Engineering, Concordia University, IEEE Circuits and Systems Society, IEEE Circuits and Systems Society Montreal Chapter, IEEE Solid State Circuits Society Montreal Chapter, IEEE Montreal Section

Image saliency is a measure of “distinctiveness” of an object or a region in an image with respect to the background surrounding the object.  Typically, these are the objects that capture our attention while viewing an image. This subjective notion of saliency makes it extremely challenging to obtain a quantitative measure for it. The first part of the talk will deal with novel ways of extracting salient objects in a scene that try to do away with heuristics as much as possible. First, we discuss an algorithm that tries to retain the exact boundary of the detected salient objects. However, if one is interested in detecting only the bounding box of the salient object (say, for example in RoI-based image compression), we demonstrate how this can be done very efficiently using the seam carving technique.

If one is given, instead, a number of images sourced from various places and under different contexts, but having at least “something” in common, then the object common in all these images may be considered as the salient object. Extraction of this object in all these images is known as co-segmentation. In the second part of the talk, we shall discuss how this can be achieved using an approximate maximum common subgraph matching algorithm.

Speaker(s): DR. SUBHASIS CHAUDHURI,

Location:
Room: EV003.309
Bldg: EV Buiding of Concordia University
1515 St.Catherine West
Montreal, Quebec
H3G 2W1