This shows the label layout algorithm using simulated annealing described by Christensen, J. et al. The high opacity rectangles are labels' final locations, whereas the semi-transparent rectangles were the original locations. The points are the lower left corners of the original rectangle locations and represent points that rectangles are labelling.
Simulated annealing is a random algorithm, so it's not guaranteed to generate perfect labels. As you may see, a few of the final rectangles are still overlapping. I think the number of overlaps can be reduced using the same compute time by focusing the randomization on regions of high overlap.
The specification to generate this visualization is shown on the right
Christensen, J., Marks, J., and Shieber, S. (1995). An empirical study of algorithms for point-feature label placement. ACM Transactions on Graphics (TOG), 14, 203-232.