You might use the image stitching approaches based on local keypoints, such as ORB, SIFT, SURF. Here three methods useful for your topic:

The first one is performed by the same author of SIFT (Scale Invariant Feature Transform) - 

http://link.springer.com/article/10.1007/s11263-006-0002-3

The second one is an image stitching method based on SURF 

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5586723&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5586723

The third one is based on a new local keypoint (ORB):

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6126544&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6126544

More in details, the stitching is based on the matching of the local keypoints (SIFT, SURF, ORB) in the images. 

You can easily find matlab code for the extraction of SIFT and SURF on the web. 

Good Luck!

Alessandro 

Hi Dear

I think the following project could be useful, please follow the next link

http://www.codeproject.com/Articles/95453/Automatic-Image-Stitching-with-Accord-NET

Thanks

@Alessandro Bruno and Faisal Mohammed :

Hi, 

Thanks for your answers. But wouldn't those methods be computationally expensive?

I mean finding out key features only will take much more time with blending operation it will be computationally expensive. Actually, I am trying to define something more simpler operation so that it will reduce computational time as well as power .

Regards,

Pankaj Wasnik