Image compression by redundancy reduction
ICA'07 Proceedings of the 7th international conference on Independent component analysis and signal separation
Hi-index | 0.00 |
Redundancy reduction as a form of neural coding hasbeen since the early sixties a topic of large research interest.A number of strategies has been proposed, but the onewhich is attracting most attention recently assumes that thiscoding is carried out so that the output signals are mutuallyindependent. In this work we go one step further and suggestan algorithm that separates also non-orthogonal signals(i.e., "dependent" signals). The resulting algorithm isvery simple, as it is computationally economic and basedon second order statistics that, as it is well know, is morerobust to errors than higher order statistics, moreover, thepermutation/scaling problem[10] is avoided. The frameworkis given with a biological background, as we avocatethroughout the manuscript that the algorithm 拢ts well thesingle neuron and redundancy reduction doctrine, but it canbe used as well in other applications such as biomedical engineeringand telecommunications.