Wednesday, July 13, 2011

COSI 2011 - part 2

Hey everyone..!

Yesterday and today were packed. We met lots of people, heard lots of talks, had lots of discussions. At smaller meetings like this there is plenty of opportunity to meet people and chat. The breaks tend to be well spaced, and then yesterday there was a networking lunch and an evening reception. So it’s been interesting, intriguing, sometimes highly entertaining, and finally exhausting! :) There is still another morning’s worth of events to go. But most of the conference is almost over. Has been such good fun!

My colleagues - Prasanna, Jorge, David and Kathrin had a talk on design for better depth estimation. It was very well received. I think this was a really good meeting to present work from our group.

Now let me give you all an update on some of the interesting talks we saw past few days. This post is going to get terrifically newsy now.

Ravi Athale and David Brady gave some interesting talks dedicated to the memory of Dennis Healy and his vision for sensors and imaging. Ravi proposed all sorts of ideas and visions for the future. History and vision go very well together.  Certainly got everyone thinking.

There were many talks on Holography. They started out with Oscan’s group discussing LUCAS, an incoherent holography technique to image cells. Some of his results use prototype attachments for compact cell phones.

George Barbastathis’s group showed much work on 3D optical imaging. He gave a very interesting talk on quantitative reconstruction of phase objects, visualizing 3D flow, bubbles, etc. using digital holography including ideas for compressive sensing.

My own (earlier) group from Rochester, Jim Fienup’s research lab, showed results by Abbie Tippie on lensless synthetic aperture digital holography for gigapixel imaging. This work involves generating a digital hologram of an object and building up the hologram by scanning the detector to effectively obtain a larger aperture and more resolution. Abbie talked about sampling conditions, propagation approximations, defocus, aberrations, drift and other mosaicking errors and how to handle them all.   

David Brady’s group also collaborated on this project. They too showed results, by Sehoon Lim, on the corrections they have implemented for their system. Both groups had slightly different implementation issues. I love that this collaboration got so much research work done on that front.

Optical superresolution, as some of you may know, is the area of my thesis research. So I love work on superresolution. Zeev Zalevsky was at COSI today and gave a great talk on recent advances in superresolution imaging. This was a very nice review style talk with tons of info. He had so many slides to show. Even a 40 minute slot is sometimes not enough. I’ll put in a link here if I find his material or paper somewhere online.

Alden Jurling from the Fienup group talked about some interesting approximations that they found worked for them and helped to speed up broadband phase retrieval. He actually blurred some of his data to get rid of high frequencies (treating them like noise), which allowed him to quickly zero-in on a reasonable estimate. Speeding up algorithms is often critical for practical implementation. So it’s good to see them explore the regime of speed versus accuracy to try and find a balance.

Sam Thurman showed a useful method to estimate the OTF of an imaging system – phase and magnitude, using a binary Siemens Star Target. The cool thing is that his technique works even for undersampled images. This is probably because the Siemen’s Start Target has repetitive identical lines which, in my understanding, gives him multiple “measurements” and allows him to get around the sampling limits.  

Fredo Durand gave a very nice talk on computational photography, briefly describing the work on light fields, coded imaging, feature matching, SIFT, blind deconvolution and superresolution as done in the computational photography community.  

His talk was followed by another great talk by Edward Adelson where he demo-ed as well as talked about an elastomer they have developed, called GelSight, that easily deforms when touched by a feather, brush, thumb, even bubbles. This deformation can be seen on the other side of this skin-like Gel and imaged in 3D using photometric stereo. The images were impressive.

Ref: and

Ramesh Raskar’s group had an interesting talk on their picosecond camera – a streak camera that can capture picosecond phenomena, as long as they are repetitive events, using pulsed light and strategically sampled frames. They demonstrated images of light traversing a bottle, bouncing off of objects, etc.

Ed Dowski gave an interesting perspective on how optics may one day be manufactured like chips. Scalable, cost effective and easy to design.

Oh, and there was much work on CASSI and other such spectroscopy techniques, a lot of good talks from the Brady and Gehm groups.

It’s getting late now. But there is still more I want to note here. I’ll write again soon. 

Till then, cheerio and thanks for reading!

No comments:

Post a Comment