Prize winning class team project for improved image processing

The project entails investigating a recent paper and both reproducing and extending the research.

grad students eecs 556 Enlarge
Prof. Jeff Fessler, Madan Ravi Ganesh, Leyou Zhang, Adeline Hong, Farhan Baqai (Apple)

An interdisciplinary team of three graduate students earned prizes (Apple iPad Air2’s) in the graduate level course, EECS 556: Image Processing, thanks to the sponsorship of Apple. The course, taught by Prof. Jeff Fessler, covers the theory and application of digital image processing, which has applications in biomedical images, time-varying imagery, robotics, and optics. Students investigate a recent paper and both reproduce and extend the research.

The winning project, Object boundary detection using decoupled active contours, by Madan Ravi Ganesh (MS student in EE:Systems), Adeline Hong (PhD student in BME), and Leyou Zhang (PhD student in Physics), confirmed the findings of the original paper, Decoupled Active Contour (DAC) for Boundary Detection, Mishra, et al., which focused on the issue of detecting the boundary of the object of interest and its background in a given image.

The algorithm created to accomplish this boundary detection was proven to be superior to existing methods, and is illustrated below:

flow diagram of algorithm Enlarge
Flow diagram outlining each stage of the DAC algorithm and its corresponding output. Blue pixels indicate aggregation of sampling points after importance sampling.

The Michigan team was able to extend the DAC method to incorporate color information from the images while also enhancing DAC’s performance, as shown below:

algorithm comparison Enlarge
Explore:
Graduate students; Jeffrey Fessler; Signal & Image Processing and Machine Learning; Student News