Computational Imaging: Finding Structure from Randomness
Prof. Salman Asif, Department of Electrical and Computer Engineering, UCRWe face two broad challenges as we design the next generation of intelligent and interconnected devices:
On one extreme, these systems will collect an enormous amount of data from a multitude of sources and require low-complexity, versatile algorithms that can make sense of all the data.
On the other extreme, certain physical or system constraints on sensing devices, such as cost, energy, time, or size will limit us to imperfect or incomplete observations. In this talk, I will discuss my research along the latter extreme. I will present some of my work on computational imaging in which co-design of sensing hardware and reconstruction software enables new imaging capabilities under different physical constraints. I will discuss how we can design simple sensing systems that preserve the information-of- interest, which we can then recover via computational algorithms that exploit low-dimensional structures in the signals. I will first present my work on FlatCam: a coded mask-based design for lensless imaging that can enable cameras with extremely thin form-factor and flexible geometry. I will then present some work on phase retrieval for long-distance imaging and self-calibration in dynamic MRI, both of which boil down to the computational problem of recovering a rank-one structure from linear observations.