Combine pictures by using SIFT

Progress:

STAGE 1:  FINISHED

Translate and understand Lowe's paper 

STAGE 2: FINISHED

Compare sample code to his paper, debug and understand his code

STAGE 3:   UNDER PROCESS

Write my own program base on his sample,see why and how it works.

STAGE 4: NOT STARTED YET

Improve his code and add SURF-SIFT algorithm to the sample code.

Detail:

Papers in STAGE 1:

[1] D. G. Lowe. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 2004.

[2] Y. Ke and R. Sukthankar, Computer Vision and Pattern Recognition, 2004 

Code used in STAGE 2:

Since there is no MATLAB in my computer lab. I will use Rob Hess's code instead of David lowe's.

The code I started to learn was programmed by Rob Hess.

OpenCV and GSL library must be installed to compile the code.

Original code can be downloaded here: http://web.engr.oregonstate.edu/~hess/

Program in STAGE 3:

Main feature of the program:

Find relevant areas in each picture, combine them to one image.

    I expect the final program can combine several images captured by cell phones or cameras to a 360 degree image like Google street view does.

Current progress:

     Combine several images to one:

 img1:

img2:

img3:

img4: 

After combined:

temporary output: