Distortion invariant optical pattern recognition

Pattern recognition involves assigning an unknown signal to a specific class. This is a difficult task because the set of signals associated with a class can vary widely in the Euclidean distance sense. For example, an image is effected by factors such as: perspective changes, lighting conditions, imaging environment (e.g. intervening clouds), and occlusions. All of these factors act in concert to generate a wide range of possible images for the same object. These variations are referred to as distortions. For simplicity, we will only discuss a two class pattern recognition problem, where a signal is classified as either target or clutter. The extension to multiclass problem is straight forward