S. Matsuzawa, N. Ishihara and S. Komatsu: Self- Organizing Algorithm for Blind Deconvolution, Extended Abstracts (Optics Japan 2004) , The Optical Society of Japan, 162-163 (2004)

A self-organizing algorithm for blind deconvolution is proposed. This algorithm consists of alternatively operating two quasi binary Hopfield neural networks (HNN). As the state of neuron is successively updated, the corresponding cost function becomes the lowest, and the restored image would be optimized. The feasibilty of the algorithm is examined by computersimulation on a blurred image with 15x15 pixels. We also use the Fourier iterative algorithm to improve effectiveness.