16 Mar 2016

For my visualization, I wanted to study areas of visual interest and disinterest in images. Originally IĀ intended to use satellite maps, and try to create maps of the interesting and disinteresting parts of a maps of the USA. However, I eventually decided to use structured images. I also decided to play with seam carving, as algorithm for removing ‘seams’ (paths of pixels in an image) of low interest.

I was interested in exploring interest and disinterest in pieces of film, since these are structured images, and I though it might be interested to apply the results I was getting to moving sequences of images.

Here are some of my output clips: The first 2 are based off some film with Al Pacino in it, and the second is part of the opening shot from Snake Eyes (19898).

This is the Pacino shot with all the visually disinteresting pixels removed:

The Pacino shot with all the interesting pixels removed:

And the Snake eyes shot: for this I was experimenting with a (broken) algorithm for filling in the spaces left when the interesting pixels were removed:

My code’s on GitHub

Overall, I like the outcomes of my experimentation, although I don’t have a concrete ‘piece’ to show. If I had had more time, I would have probably developed a more sophisticated filling algorithm and use that to create long modification of a piece of video.