An information theoretic metric for identifying optimum solution for normalized cross correlation based similarity measures

Similarity measures such as normalized cross correlation (NCC) are widely employed for applications such as pattern recognition and/or template matching which are commonly used in image registration. This approach, however, is not immune to noise variations present in the images especially in case where multiple bands of interest are dominated by both system and external noise present in the sensor´s field of view. Thus noise can influence the calculation of correlation coefficients and produce erroneous results during template matching. This work proposes a metric which identifies the best NCC coefficient value or values in case of a spectral data cube, for optimized application of similarity measures for template matching.