Project abstracts
Expansion-compression
Variance-component Based Sparse-signal
Reconstruction from Noisy Measurements
Kun Qiu
Abstract:
In this project, we propose an expansion-compression variance-component
based method (ExCoV) for reconstructing sparse or compressible signals
from noisy measurements. The measurements follow an underdetermined
linear model, with noise covariance matrix known up to a constant.
Compressible or sparse signal structure is imposed by defining high-
and low-signal coefficients, where each high-signal coefficient
is assigned its own variance, all low-signal coefficients are assigned
a common variance, and all the variance components are unknown. We
apply the proposed method to reconstruct signals from compressive
samples, compare it with existing well known approaches such as CoSaMP,
GPSR, l1-magic, etc. and demonstrate the performance via numerical
simulations.
Estimating Sparse Contour Deformations using Compressed Sensing
: Applications to Deformable Contour Tracking
Samarjit Das
Abstract:
In this project, we attempt to explore the possibility of using
compressed sensing in order to estimate temporal contour deformations.
Thus, starting with the initial contour, our ultimate goal is to
develop a compressed sensing based deformable contour tracking
algorithm. We assume that the associated deformation signals are sparse
in frequency domain and hence we can recover them from a very
small number (relative to the full contour length) of random
measurements. Our algorithm takes the initial contour information and
reconstructs the contours corresponding to the successive time
instants from random observations. The observations are considered to
be randomly sub-sampled version of the true deformation signal. We have
used a nearest edge-based search method for the observation extraction.
We demonstrate the working of the algorithm for multiple examples, both
simulated and real-life sequences; including how the system could
possibly work as a deformable contour tracking algorithm. The
methodology developed in this work has potential applications in many
areas including deformable contour tracking over biomedical image
sequences which are often corrupted with noise and
occlusion/clutter scattered randomly all over the contour. In that
scenario, we can extract observations using the points which are not
corrupted (also assumed to be randomly scattered over the contour) and
'robustly' reconstruct the entire contour using our technique.
COSAMP and Sequential COSAMP
Fardad Raisali
Abstract: This project investigate the performance of the
OMP(Orthognal Matching Pursuit) and Cosamp(Compressive sampling)
in the reconstructioin of
the sparse signals. We use MR images here as our signal.Also we would
like to test the sequential MR image reconstruction which the
main aidea is using
previous estimate to improve current one.By this approach we expect
an increase in the speed of the algorithm.