Final Project Directions
CS 674 Computational Photography

Proposal Requirements

Please provide a description of the project you intend to do. It does not need to be a lengthy proposal. I would expect something around half a page. If your project is inspired by a paper please refer to it. Please provide the starting point, which means all the resources you already have (e.g. source code and data), and describe what is it that you need to add to achieve your goals. If you are not going to implement a paper entirely, indicate which part you are going to take care of. I will offer recommendations as to whether you are attempting to do too much or too little. As a rule of thumb, I would expect that for a one-person-project, the level of effort should be around twice as much as the level of effort of one assignment.

Project Requirements

You are required to chose a topic and develop a project based on it. The general way to do so is to pick a paper that you are interested in, and develop your proposal based on it. You could chose among the papers we covered or mentioned in class. You could use a paper in the current reading list, you could also look trough the reading lists of similar classes in Computational Photography taught at other institutions, or you could look for papers on your own in the computer graphics, computational photography, and computer vision conferences. You can also get inspiration from the final projects from computational photography classes of previous years in other institutions. As stated above, if a paper is very rich, you may want to implement only portions of it, or a simplified version. If you want, you can also expand one of the assignments you already did, but you need to show that you can make extensive improvements.

A good strategy to propose something compelling is to apply ideas developed in class or from a computational photography paper of your choice, to your own research (provided that it is not done in a trivial way). You could also compare different approaches (for instance one approach developed in one of the assignments with another one). You could develop your own extensions to a paper.

The projects can be either individual or you can form teams of two. Obviously a team of two is expected to make double the effort. Teams of two are strongly encouraged!

Final Project Presentation

Deliverables

You should deliver the project proposal to me via email as noted above by November 20.

You should deliver the project package by December 8.

The project will be graded out of 100 points. There will be no extra points since you are defining the specifications of your project. Obviously, the more work you do and the highest the likelihood to secure an A.

To turn in your assignment, you will need to deliver a zipped file (Use these guidelines to deliver your files) containing the following:

[65 points] Code: Please supply all the code and data you used to generate your results, along with a README file to allow me to test it out.

[35 points] Web: Create a final project page to showcase your results. By now you are an expert and do not need suggestions!

• NOTE: I will not accept project proposals on panoramas!
• Eulerian Video Magnification: Implement your version of this algorithm. This is a relatively simple and very interesting way to enhance spatial and temporal image variations that are difficult to visualize with the naked eye.
• Implement the approach described in this paper to separate the direct and global illumination components of a scene.
• Search for and implement your favorite video stabilization algorithm.
• Implement the natural image matting method of Levin et al.
• Colorization. Implement your own version of the approach of Levin et al. or use their code and compare their approach with another one of your choice that you have to re-implement, e.g. the one in Image Analogies.
• Implement an algorithm for video abstraction, like this one.
• Create fake miniatures by simulating tilt shift photography.
  
• Gradient Domain fusion: Implement an interactive tool that allows you to select, copy and paste of a portion of one image onto another image. Test the mixed gradients approach and propose your own extensions to this process.
  
• Image Analogies: Implement your version of the algorithm; propose a new kind of analogy that was not proposed before; test several types of analogies; ...
• Texture Synthesis: Implement your version of a texture synthesis approach, or compare two approaches. Show examples of texture enlargements and hole filling; ...
• Hole filling: implement a data driven hole filling approach. This could be either the approach of Criminisi et al. or another one, based on large Internet data collections, such as the scene completion approach of Hays and Efros.
• Implement the non photo-realistic rendering approach of GradientShop.
• Super-resolution from a single image. Implement the approach of Glasner et al. to recover the high frequency content of an image.
• Creation of David Hockney Joiners. The artist David Hockney has created images, known as "joiners", that consist of a collection of photos of scene, taken from a wide range of viewpoints, arranged in a collage to give a photo-collage yielding a non-realistic but compelling representation of the 3-D scene. Creating these currently involves a lot of manual alignment and artistic skills. The project would involve creating a tool that would let a user create these images much more quickly, while retaining artistic control over the process. This is a software-only project which an artistic and human-computer-interface flavor. (Project idea borrowed from Rob Fergus.)
• Hot or not. This project explores ways to parametrize our notion of beauty. First, a large collection of images of people from websites such as Hot Or Not should be collected, along with metadata such as viewers ratings, hair color, eye color etc. After alignment of the faces, you should try and build simple models that allow a user to traverse the space of faces in a logical manner, e.g. blonder hair, prettier, thinner face etc. Some machine learning may be required, so is more suitable for someone who took the Pattern Recognition class last semester. (Project idea borrowed from Rob Fergus.)