We propose a novel algorithm for the independent component analysis that reshapes a scatter diagram or the joint probability density function of the pixel values of the mixed signals. The algorithm outperforms the conventional algorithms in speed and accuracy. The detailed procedure of the algorithm is described and numerically evaluated. The present algorithm can separate two independent images from two mixed images. By linear transformation of the two mixed images, shearing and rotation are performed to the scatter diagram. After the conversions and normalization, the basis vectors of the independent signal become orthogonal and match to the coordinates. In the implementation, no visible degradation and corruption were found in the final images. In additions, the proposed algorithm are compared with two conventional algorithms; [G. Burel: Neural Networks 5, 937 (1992)] and [E. Oja: Neural Comp. 9, 1483 (1997)]. The calculation time of the proposed algorithm was shorter than that of conventional algorithms. The analyzing error, which numerically evaluates the difference of pixel values between original images and retrieved images, was the least for our algorithm. .

Keywords: Independent Component Analysis, signal processing, signal mixture, blind signal separation, scatter diagram.

Proc. SPIE, 3740,142-145 (1999)