Blind Image Source Separation based on MMCA using Dictionary Technique
Anju Thomas, Durai Sugumar · 2013
In real world image processing applications, blind source separation process has found to be very important. The goal of blind source separation is to separate and recover the image sources from their instantaneous mixtures without knowing the mixing parameters. Sparsity found to be very useful for the separation process. Each of the sources needs to be sparsified by using some known transform. For the case of multichannel image source separation, if a prior knowledge about the source sparse domain is not available, then the existing algorithms will fail to recover the sources. In this paper, the problem is solved by using a new algorithm that improves the quality of source separation by using dictionary learning technique for multichannel observations in both noisy and noiseless situations. The dictionary is learned using single value decomposition algorithm and the result shows that the recovered image sources are more accurate.