Nevatia - 5 
Figure 5. Outlines of two detected objects from Figure 2(a) 
a) Complexity : The approaches outlined require examining relationships between different sets of 
observed features. If we need n features out of a total of m features, the computational complexity 
could be 0(m n ). However, we can control this complexity in many ways. We can limit the 
neighborhood in which a search for grouping features is conducted based on the maximum size of 
expected objects, distances between the tested features as some limits on size of desired objects. 
We can also organize the grouping process hierarchically, so that at first we look for only a small