There is the original 1993 paper by Mallat & Zhang - "Matching Pursuits With Time-Frequency Dictionaries". You might find that one useful.

Matching persuit is good but old technique. So for reconstruction you can use convex optimization. Its toolbox is available in matlab. 

http://cvxr.com/cvx/

This is the link to understand the basic of compressive sensing and reconstruction.

If you want the entire process...you can follow the paper

http://link.springer.com/chapter/10.1007/978-981-10-0767-5_2

This might be helpful:

http://www.zess.uni-siegen.de/myoffer/files/file_5_52170211052012.pdf

The following book is very good. "Sparse and Redundant Representations" by Michael Elad.