Performance of optimal trade-off and distance-classifier circular filters for rotation-invariance
We evaluated the performance of circular optimal trade-off SDF (OTSDF) and distance classifier correlation filters (DCCFs) as rotation-invariant correlation filters. Because the filters are designed using different parameters we compared the filter's performance in terms of an equivalent effect of probability of error. The use of OTSDF and DCCF filters as circular filters allows their calculation to be greatly simplified when compared to using rotated views of an object to create filters. We found that both types of filters can be used for rotation-invariant object recognition in a noisy environment. In addition, the filters generated were real-valued so they may be implemented on a variety of spatial light modulators.