Fast normalized cross correlation with early elimination condition

Real Time target tracking based on template matching has become one of the key technologies in image processing technique. This paper describes a fast algorithm based on Normalized Cross Correlation for real time target tracking in a video sequence. The Normalized Cross Correlation is efficiently implemented in practice with the concept of Box Filtering and Integral Image to eliminate redundant computations. But NCC-BF and NCC-II still have lot of redundancies in exhaustive template matching process. This paper proposes an upper bound criteria called as Early Elimination Condition, which states that the calculated optimum correlation score at current position is lower than the maximum correlation score obtained in previous position. This condition accelerates the matching process. In this paper, it has been proved that if this condition is verified, then the template can proceed with next reference position without executing the rest of operations in current position. Hence, the redundancy in NCC based object tracking can be efficiently reduced by our proposed elimination condition described in this paper. The performance of the upper bound criteria is established by tracking various objects captured from large set of video clips.