Alignment of Elastically Deformed Electron Microscopy Images The Computational Connectonomics group is trying to take huge volumes of detailed SEM images from mouse brains and use computational methods to learn about neuron interconnections from the images. There are many steps involved in this process the first of which is image alignment. Essentially because the brain slices are extremely thin you end up a detailed 3-D visualization of the mouse brain such that there are layers of 2D images that are overlaid. I plan on finding a fast effective way to find identical features between the two overlaid images made more difficult because the transformations between images are nonlinear. Our plans to approach this problem include utilizing optic flow algorithms and convolutional neural networks.
I really wanted to work in an area that would be able to integrate my interests in machine learning and low level programming / performance engineering and this project seemed perfect for my goals.