Wavelet Transform

The original image is passed through three wavelet transforms shown here as horizontal, diagonal and vertical. The icons on the right side of the transforms represent the coefficients that represent the image. We routinely discard every other coefficient. A 1/4 resolution representation of the images is then passed through the transforms again, and half of the coefficients are discarded. Finally a 1/16 resolution of the image is passed through the transforms. Now correlations are found between the different transforms at each resolution. Coefficients that are not correlated are discarded because the likely represent spurious activity. The number of coefficients in the final representation is smaller than the number of pixels in the original image. Processing power is a function of the number of pixels that must be processed, so this waveform transformation is a power-saving operation.

Prioritization

Think of the red and green surfaces as flexible "blankets" that descend upon the representation of the two images below, one from the thermal camera, the other from the visible camera. As these blankets descend, they deform to follow the high regions of the images. These high regions represent the areas of highest information content. The blankets now contain the addresses of the regions of each image with highest information content. The values in the blankets for each image are compared. The the regions of highest information content are sent to the monitor.