CS 7495 -- Computer Vision
Project IV -- Coin Hunt
by Anna Shleyfman

=====================================================
Source code:  coinHunt.m, findLocMax.m, drawCircle.m, colorCoins.m.
=====================================================

Summary:

This program uses Houph Transform to detect coins in the image.  Originally it used to search 20 radii in order to find the coins.  Unfortunately, that approach caused the program to run for a very long time (due to the nature of Matlab software package).  So, I determined the radii of each coin in the picture, and give the radius of the coin as an input parameter.  The use of only one radius reduced the Houph Transform computation (coinHunt.m) time to about a minute. Also, I broke the program into two parts: Hough Transform computation, and detecting the position of coins by evaluating matrix, generated by coinHunt.m. This matrix holds the "voting" information for each pixel. This second part of the program is called findLocMax.m, which stands for Find Local Maximum.
 
Code:


Output:

The following pictures were outputted by my program. The program was run twice: first, to detect quarters, and after that, on the produced result to detect cents.
Hough Transform for quarters
Hough Transform for cents
Here, the output from the Houph Accumulator array, are scaled by the highest vote for the center of the pixel.
Original Image
Detected Coins
As you can see the program found all of the coins: there is a thin line around all of the quarters.


==================================================
If you have any questions please e-mail me at:  shleyf@cc.gatech.edu