Joint Bilateral Filter (JBF)

作者 PrintDebug

In this project, the main motivation is just to accomplish the homework in computer vision. I will capture some interesting parts of this project and display them below. Thanks to my attitude of writing homework seriously, it is still quite easy to review JBF now.

At first, I introduce what is the JBF, this filter contains two weights which are obtained by spatial kernel and range kernel, respectively.

The output is shown as follows:

And we can see some good result about this filter.

We use a 3 channel picture as input, and its grey picture as guide picture. we can see the result about this algorithm.

In short, joint bilateral filter can make the image smoother. But I think the point to think carefully is, what are the benefits after changing the picture? What are the disadvantages of not having to guide pictures?

My answer is if you have a good guide pic, this algorithm will be much more stable, casuse the center have to consider the piexl around itself and this consideration is reasonable. However, if the guide pic is not a good pic, the result is not as good as bf.