Detecting shapes

So, we have seen how to detect edges; however, this process is a pixel-by-pixel process answering the question of whether this pixel is an edge or not. Moving forward, in shape analysis, we would need more concrete information than just the edge test; we will need a better representation.

For example, if we have a picture of a box and we did the edge detection, we will end up with thousands and thousands of edge pixels; however, if we tried to fit a line to these edge pixels, we get a rectangle, which is a more symbolic and useful representation.

Understanding the Hough line transform

There are many ways to fit a line through a number of points, and Hough transform is considered an under constrained method, where we use only one point to find all the possible lines that can go through this point and we use another point to find all the lines that can go through it too, and we keep doing this for all the points that we have.

We end up with a voting system where each point is...