Illumination invariant adaptive joint transform correlator

摘要

Illumination functions can be modeled as multiplicative noise. Therefore, an input scene for pattern recognition is commonly corrupted by a mixture of additive and multiplicative noises. Existing correlation filters assume that multiplicative noise is either a constant factor or wide-sense stationary noise We propose a point-wise preprocessing of an input scene and consequent optimum correlation filter design. This procedure takes into account knowledge of illumination function. Adaptive joint transform correlator architecture is used. Computer simulations results are provided and discussed.