An exciting new method is presented to enhance visual images which occur in numerous biomedical imaging applications. The procedure employs some methods from nonlinear dynamics and has been shown to significantly improve object recognition and signal to noise ratio in the processed image. Tests were initially conducted on one dimensional signals to demonstrate that the quality of information could be amplified by a factor greater that one if certain computational processes were applied to the input signal.

    The results from one-dimensional signal tests showed that a substantial amplification of signal to noise ratio could be obtained for certain elementary signals. This seminar will then generalize this concept to multi dimensional signals including visual images. The relationship of this procedure to statistical decision theory is described to better understand how the procedure assists in improving decision-making for key object identification. A further generalization of this method is then given with respect to the analog of hyper-spectral imaging to the biomedical imaging community.

    The key concept of fusing the results will be presented. Results from ongoing work will be presented on visual images that are significantly confounded by noise and other anomalies.