Homework answers / question archive / Before you start For your projects you will use github

Before you start For your projects you will use github

Writing

Share With

---

Before you start

For your projects you will use github. You can use your own account and create a

private repository for the project.

To use the Eigen::Vector3d, add -I/usr/include/eigen3/ to your compilation flags

and include .

CImg is a great library for image reading and writing. Here is code that simply reads

and writes an image.

To make sure you do not have last-minute problems, you must commit and push an

updated README.txt file to your repository by Tuesday, March 22. You should

replace the default entries in the readme file with your own information. The default

is:

gl_login, full_name

I would replace this with:

adamb, Adam Bargteil

We will be using this data to initialize our grades spreadsheat, so it is important that

you use comma separated value format with the correct number of entries in the

correct order.

The Assignment

Graphcut Images

Compositing along the minimum seam between two images

Requirements

You are required to implement a graphcut algorithm as well as take your own photos

and apply your algorithm to those images to get full credit.

NOTE: You should implement max flow/min cut algorithm on your own as you are not allowed to use library for that. You should also implement either texture synthesis or video synthesis or simple panorama stitching (assuming that all images are aligned

and have identical overlaps) using Graphcut to find the seam and Poisson blending to make the seams less apparent.

Details

When compositing two images, it is often a goal to combine them along a seam (series of pixels) that has the minimal difference between the two images, hereafter referred

to as the optimal seam. For this project you will be implementing an algorithm that finds the minimal seam between two images.

If you've taken a graphics course at Brown, you've probably heard of Seam Carving.

You may have even implemented it. That algorithm uses dynamic programming to find the minimal seam of an image along a specific dimension, and then remove it.

Unfortunately, that doesn't generalize as well to finding seams that are not along a single dimension.

An algorithm called Graphcut (website), was created to address this problem. The algorithm works by formulating the image as a connected graph with edge weights