Unsegmented target classification for images degraded by uniform intensity transformations

We propose a new pattern recognition technique that is invariant under any arbitrary uniform intensity transformation. Any image that undergoes an intensity transformation can be written as a linear combination of its gray level slices. The classification problem then consists of determining if an object is part of the vector subspace created by those functions. This is done by means of nonlinear combinations of multiple correlations. We also propose a way to control the discrimination sensitivity. Because this method uses mainly correlations it can be implemented optically.