Mutual information-based image template matching with small templatesize

Mutual information plays crucial role of similaritymeasure in some applications such as image registration and imagetemplate matching. Therefore estimating the joint probabilitydistribution of underlying image or template is the main problem.Non-Parametric window (NP) method considers the images ascontinuous two-dimensional signals and results an appropriatejoint probability distribution. In this paper we employ a triangledistribution with a large support instead of uniform distribution inthe original NP. This gives a more precise estimation of jointprobability distribution. As a result the compared mutualinformation is more robust and reliable. Experimental results showthe superiority of the proposed method in image template matchingwith small window size.

Keywords:

component , Template matching , joint probabilitydistribution , mutual information , Non-parametric windowmethod