THE MEDIUM OR THE MESSAGE? DEALING WITH IMAGE SPAM
Catalin Cosoi · 2006
However, the false positive rate of such techniques is rather high. Of course, when you’re just searching for pretty pictures, more is better. In an anti-spam solution, on the other hand, false positives are a serious problem. The types of distance functions commonly met in old CBIR systems are: histogram Euclidean distance, histogram intersection distance and histogram quadratic (cross) distance. Of all these options, the most appealing for an anti-spam engine would be the histogram intersection distance.