Reputation: 437
I'm writing a script to auto-detect some circles on a set of photographs. I've probably read every question about HoughCircles on Stack and the recommended approach generally seems to be to brute force both the radius interval and the accumulator threshold (i.e. param2). The range for the radius loop can generally be estimated from other geometric properties of the image, but what is a good approach to limit the search range for param2 automatically?
If anyone has any other tips for optimizing this brute force approach I'd also appreciate it.
Upvotes: 0
Views: 122
Reputation: 1915
It is not about param 2. It an accumlated effect to get things right.
Let take the following example. With the input
we can get different feature map according to a different parameter in other function like canny or other segmenation output
This feature is only good for line. whatever you choose for param2 the left circle wont come out
Here is another setting, by have it here, you can have the circle detected at param2 = 110
You see there is no perfect or optimized search range for param2. But the way you approach the problem matters.
Upvotes: 1