Thursday, 27 September 2012

Breathtaking Images from a Computer Science Assignment

It’s true. I found these from Dr. Vincent Pegoraro’s Assignment no. 4 for CS 7966. They are called Hough Transforms.

Ethereal, I know. Not only do they look absolutely gorgeous, they are also an useful feature extraction technique for identifying lines or curves in an image. It works like this:  imagine all the pixels in an original image (a) are sorted by a sift (b) such that only pixels that lie on a certain line can pass through. These pixels are then collected in a designated spot on the transformed image(c). Repeat this for all the lines you can possibly draw in this 2D space – every direction and every position. What you get at the end is a Hough Transform (c).
Fact: Hough Transform sorts image pixels into a gorgeous picture.

In case you were wondering, here are the original images for the beauties above.
Bright spots (intersections of curves) on the transformed image indicate how many prominent lines there are in the original image. For example, there are 7 bright spots in the 2nd transformed image, for the 7 edges forming the T shaped object (two of them lie on the same line).
Math is fun, and pretty.

1 comment:

  1. Your Blog is very helpful for me. Your navigation & structure of website is quite well. We also serve same domain to know more about us follow us on Assignment web Thank you.