Oregon State University

We’d like your feedback: Calendar User Survey – Event Creator Survey



Event Details

MS Final Examination – Sharath Kumar Dhamodaran

Friday, April 5, 2013 3:00 PM - 5:00 PM

Pattern Discovery in Noisy Images
The Focused Ion Beam (FIB) tool is a versatile instrument for nano-matching in circuit editing. Circuit editing is one of the most important steps in the design of an electronic circuit on a chip. Circuit editing can be improved by imaging of silicon plates and analyzing the resultant images. However smaller features in FIB produce lower signal-to-noise (S/N) ratio in images, and thus render visual feedback increasingly difficult.

For this Master thesis, we focus on analyzing a set of Focused Ion Beam grey valued images. They are the result of temporally averaging over a span of time. Our goal is to reduce the noise in images, and apply an appropriate filter, for feature extraction. This is expected to allow us to see the significant features of the silicon plates. The main challenge is that the noise and its level are unknown. Also the significant features of the silicon plate are unknown. Therefore the problem that we face here is discovery of an unknown pattern in images, under unknown noise.

We make the assumption that noise is a superposition of a Gaussian noise and Salt-and-Pepper noise. We seek the significant features using a least squares optimization. To account for possible variations in the plate size and position during the imaging process, we also subject the images to affine transformations. Thus, in the least squares optimization, we solve for the latent affine transformation and then recover the most significant pattern across the images. The improved images enable the tool user to recognize critical features more reliably.

Major Advisor: Sinisa Todorovic
Committee: Xiaoli Fern
Committee: Huaping Liu
GCR: Arijit Sinha 

Kelley Engineering Center (campus map)
Nicole Thompson
1 541 737 3617
Nicole.Thompson at oregonstate.edu
Sch Elect Engr/Comp Sci
This event appears on the following calendars: