Soft Scissors
Spent some time looking over 2007 SIGGRAPH papers...
One that interests me is Soft Scissors: An Interactive Tool for Realtime High Quality Matting. Matting is an area around which I have some experience and passion. When you're extracting an object from an image, hair always presents the biggest challenge, which is why the examples on the Soft Scissors page focus on it.
Almost 10 years ago, I created this tool. At that point in time, the problem was utterly foreign to me. The leader of the team showed me an example of a similar tool in action and asked me if I thought I could do the same thing. I told him I'd give it a try and spent a week on the problem.
I remember thinking "Let's see. They want me to separate the region of interest into two groups, because the groups are different. How do we determine if two groups are statistically different? Student's t-test is one way. So let's separate the groups at the point where the t-test says they're the most different." Later, I realized that I'd rediscovered a something similar to a common linear discriminant.
One that interests me is Soft Scissors: An Interactive Tool for Realtime High Quality Matting. Matting is an area around which I have some experience and passion. When you're extracting an object from an image, hair always presents the biggest challenge, which is why the examples on the Soft Scissors page focus on it.
Almost 10 years ago, I created this tool. At that point in time, the problem was utterly foreign to me. The leader of the team showed me an example of a similar tool in action and asked me if I thought I could do the same thing. I told him I'd give it a try and spent a week on the problem.
I remember thinking "Let's see. They want me to separate the region of interest into two groups, because the groups are different. How do we determine if two groups are statistically different? Student's t-test is one way. So let's separate the groups at the point where the t-test says they're the most different." Later, I realized that I'd rediscovered a something similar to a common linear discriminant.
0 Comments:
Post a Comment
Subscribe to Post Comments [Atom]
<< Home