Translation, scale and rotation invariant object recognition
Igor S. Gruzman, Vladislav G. Nikitin · 2003
Over the last years the problem of pattern recognition under various geometrical transformations is very important. In the case of affection scale and rotation transforms, well-known correlation-extreme algorithms are not efficient. The invariant object recognition methods of translation, scale and rotation are suggested. Root-mean-square method is more computationally expensive, but it provides good recognition when alien objects have different shape with object size error up to 25%.