Absolute Error minimisation-based Compressive Sensing Reconstruction

Wattanit Hotrakool and Charith Abhayaratne


We have proposed a new fast compressed sensing reconstruction using the least square method with the signal correlation. It is well known that the complexity of l1-minimisation is very high and is undesirable for many practical applications. The least square method, on the other hand, has a much lower complexity. However, the least square method does not promote the sparsity of signal and therefore cannot provide acceptable reconstructed results. The main contribution of this work is to show that by exploiting signal correlation, the reconstruction error of least square is greatly improved.

We have also presented a method for online estimating suitable references for video sequences using the running Gaussian average. This method can provide robustness to video content changes as well as reconstruction noise. The experimental results demonstrate the performance of this method as superior to those of the state-of-the-art l1-minimisation methods. The results are comparable to the lossless reference reconstruction approach.

Experimental results show that the performance of this approach is comparable to the state-of-the- art algorithms, whilst having a much lower complexity. It also shows that this method can be applied to both sparse and redundant signal reconstruction.


1: Low activity sequences

1.1: Akiyo

1.2: Claire

1.3: Highway

2: Medium activity sequences

2.1: Coastguard

2.2: Hall

2.3: News

3: High activity sequences

3.1: Foreman

3.2: Ice-skate

3.3: Silent


1 W. Hotrakool and C. Abhayaratne, “Fast compressed sensing reconstruction using least square and signal correlation”, in Proc. IET Intelligent Signal Processing (ISP 2013), 2013.
2 W. Hotrakool and C. Abhayaratne, “Running Gaussian Reference-based Reconstruction for Video Compressed Sensing”, (to appear in) Proc. IEEE International Conf. on Acoustics, Speech and Signal Processing (ICASSP 2014).


1 ICASSP 2014 poster