Conclusion
The mixed-mode design technique can be quite useful for implementing various image processing algorithms in silicon while keeping complexity and silicon area low.
Increasing the neighborhood radius of a traditional CNN can result in a topology capable of performing contrast enhancement.
Contrast enhancing gray-level images increases the margin of acceptable threshold values when converting to black & white, thus facilitating a threshold operation that can be applied to just about any facial image.
Morphological erosion and dilation can be implemented in silicon in a mixed-mode fashion and opening and closing can enhance salient features such as eyes.
The R-2R ladder DAC can be generalized to two dimensions creating a 2-D geometric distribution of voltages that can be used to perform Gabor-like wavelet correlations in silicon. This circuit also almost begs to be implemented using mixed-mode design techniques.