An Improved Shift Estimation Algorithm for Infrared Sequences

An improved algorithm for estimating the global movement of infrared sequences with fixed pattern noise (FPN) is proposed. One of the challenges in practical infrared sub-pixel motion estimation is how to obtain high accuracy with sufficient robustness to additive noise. Motivated by the fact that adaptive filter has the ability to suppress the noise, an improved sub-pixel registration method for infrared image sequences was introduced. The proposed algorithm used an adaptive filter to suppress noise, and then derived the motion estimation analytic expression by minimizing the mean-squared-error (MSE) function between the reference and target frames. Computer simulations and actual experiments' results demonstrate the superiority and adaptability of the proposed algorithm, compared with the ordinary sub-pixel motion estimation method when noise exists.