How to create panoramic image using a set of images

Question

I am a newbie in image processing and I have taken a project of creating a panoramic image from a set of images and I don't know how to get started on this .What all things I have to learn ? Thank You in advance

Answer 1

Here's a very good reference:

Matthew Brown and David G. Lowe, "Automatic panoramic image stitching using invariant features," International Journal of Computer Vision, 74, 1 (2007), pp. 59-73.

The process is basically this:

1. Extract image features at repeatable keypoints.
2. Do image-to-image matching to find out overlap areas between images.
3. Do bundle adjustment to align all the matching images.
4. Rendering tricks so that the output looks nice.

The above reference uses SIFT features, described in this paper:

David G. Lowe, "Distinctive image features from scale-invariant keypoints," International Journal of Computer Vision, 60, 2 (2004), pp. 91-110.

Prerequisites to understanding this material include:

• image gradients
• interest points (corner detection, difference of gaussians)
• descriptor types (simple image patches, SIFT, SURF)
• approximate nearest neighbor search (I'd suggest using the FLANN library for this)
• RANSAC
• linear algebra (especially homographies)
• non-linear least-squares minimization techniques (like the Levenberg–Marquardt algorithm)

Koan suggests an excellent textbook in his answer.

It's possible to use a lot of existing code to do feature extraction (VlFeat provides Matlab and C++ libraries), RANSAC, and minimization (I don't know what the best libraries are for those, though). OpenCV is also a very good computer vision library.

How long do you have for this project? I think it's quite ambitions. Based on having no experience in computer vision, I think a pretty good goal would be to find the transformation between just two images and stitch them together. You'd learn a lot just from that.