An Algorithm for Efficient and Exhaustive Template Matching

This paper proposes an algorithm for efficient and exhaustive template matching based on the Zero mean Normalized Cross Correlation (ZNCC) function. The algorithm consists in checking at each position a sufficient condition capable of rapidly skipping most of the expensive calculations involved in the evaluation of ZNCC scores at those points that cannot improve the best score found so far. The sufficient condition devised in this paper extends the concept of Bounded Partial Correlation (BPC) from Normalized Cross Correlation (NCC) to the more robust ZNCC function. Experimental results show that the proposed technique is effective in speeding up the standard procedure and that the behavior, in term of computational savings, follows that obtained by the BPC technique in the NCC case.