PPHT Results

---

• Small standard deviation leads to problems
  • Circles classified too quickly: may need a vote total only 1 or 2 more than the mean
  • Pixels from different circles may vote for the same incorrect circle
  • Artificially scaling the standard deviation during the Gaussian calculation leads to much more reasonable results, however, this approach is unfounded and indicates moreso a fundamental fault in the algorithm
• Reasoning
  • Cicles in accumulator space intersect twice as often as lines
  • Circle transform more compact than line transform
    • Incorrect "noise" votes not evenly distributed through accumulator space