On compressed sensing image reconstruction using multichannel fusion and adaptive filtering

Compressed sensing (CS) in imaging is essential in many practical situations and does need suitable algorithm for reconstruction from this highly incomplete observations/measurements. Measurements are also noisy in majority of the cases. This paper addresses CS-reconstruction of images from multiple channel data acquisition with under sampling measurements. An adaptive filtering based stochastic approximation through a recursive system is explored to reconstruct images from CS-multichannel noisy measurements using fusion technique. A simple weighted averaging is done in fusion process where the weights are calculated based on estimation of noise level. A spatial domain adaptive Wiener filter is then used to diminish the noise and reveals new features from the degraded observations. Experimental results show that the proposed multichannel fusion based CS reconstruction scheme outperforms individual channels both in subjective and objective quality measures.