Serena K.A. Thompson

Year Entered: 2002 

Degrees Received:
B.S. at University of Southern California, 2002

Honors and Awards:
CCS Fellowship, 2004-2005

Thesis Advisor: Daniel Kersten, Ph.D., Neuroscience Graduate Program

Thesis Research: 

We are working on several models of image feature saliency, that is, which parts of an image are used by areas of low level processing, such as V1, to drive first gaze and visual scanning patterns. Several approaches may be taken. Areas of 'saliency' have been suggested as areas of high entropy and thus high uncertainty; areas that will reveal the most about the entire image. It has also been suggested that saliency is context and task dependent: one will only look in certain areas of a given scene for a car, for example, or a person. What is most important in the first glimpse of an image, however, may be driven by bottom-up or top-down processing (the lines, contrast, intensity, and orientation changes, or the objects that compose the image). We are specifically hoping to find those low level features likely to be processed by V1 (by using a bottom up approach) correlated with gaze, and determine using psychophysics their relative saliency during different tasks. If an interesting relationship is found between different image features and task dependent saliency, we will follow up with fMRI studies of early visual cortex.