"The sampling theorem is wrong! ... It engenders the psychological expectation that we need very large number of samples in situation where we need very few. ...". Compressive sensing (CS) [1,2] is a recently emerged signal processing technique that opposes the conventional wisdom in signal representation and reconstruction. This novel technique enables accurate, if not perfect, recovery of signals from a small set of measurements, far fewer than that imposed by the classical Shannon-Nyquist sampling theorem, which previously believed to be highly incomplete.
This project will investigate the exciting development in CS and how it can be applied to image compression of standard 2D images. In particular, the contention is that a proper integration of CS should facilitate an improvement in the rate-distortion trade-off over existing methods.
[1] R. G. Baraniuk, "Compressive Sensing [Lecture Notes]", IEEE Signal Processing Magazine, vol. 24, no. 4, pp. 118-121, 2007. https://doi.org/10.1109/MSP.2007.4286571
[2] E. J. Candes and M. B. Wakin, "An Introduction To Compressive Sampling", IEEE Signal Processing Magazine, vol. 25, no. 2, pp. 21-30, 2008. https://doi.org/10.1109/MSP.2007.914731
Undergraduate
Implementation in MATLAB of a compressive sensing image compression/reconstruction framework and evaluation of its performance over existing schemes.
Students should have a firm mathematical background and be familiar with MATLAB. Knowledge in basic linear algebra would be advantageous, and it is recommended that at least one student is enrolled in Signal Processing (ELECTENG733).
Lab allocations have not been finalised