dhchung / image_stitching

Image stitching using homography transformation

image_stitching

Matlab code to generate stitched images using homography.

Requirements

1. Matlab computer vision toolbox
2. Calibrated images (following pinhole camera model)

Image stitching by finding the homography matrix between successive images.
Features :

1. Harris corner detector
2. SURF

Matching Method:

1. SAD : Sum of absolute distance
2. SSD : Sum of squared distance
3. NCC : Normalized cross correlation

How does it work:

1. Find features in the images and store them
2. Match features between successive images
3. Perform RANSAC with homography model to get inlier points : Reprojection error is used
4. Do NDLT (normalized direct linear transform) to get an optimized homography matrix
5. Warp the target images into base image.

How to run:

1. Name your images in consecutive order (1.jpg, 2.jpg, 3.jpg ...) (note : the consecutive images should have some overlapped region)
2. Load the images as in %imgae sample in run.m
3. Set the base image
4. run.