Image Processing in Presence of Huge Amount of Data Contamination
Md. Nurul Haque Mollah, Md. Nurul · International journal of imaging and robotics · 2010
Blind source separation (BSS) by independent component analysis (ICA) could be a popular and promising statistical technique for image processing. This paper proposes the extended minimum Beta-divergence method of local ICA as an adaptive highly robust blind image source separation algorithm. This algorithm sequentially explores local groups of original signals based on the initial value of the shifting parameter in which the observed signals follow a mixture of several linear ICA models. The value of the tuning parameter Beta plays a key role on the performance of the proposed method for blind image source separation. The performance of this algorithm is equivalent to the standard ICA algorithms if there is only one data cluster in the entire data space, while it shows better performance otherwise. Our experimental results also show that the proposed method is able to extract original image signals in presence of huge amount of data contamination.